2025-01-16_uttam_kumarans_zoom_meeting_4de73a5e

Meeting Title: Uttam Kumaran’s Zoom Meeting Date: 2025-01-16 Meeting participants: Brian, Uttam Kumaran

---

WEBVTT

1 00:00:40.430 ⇒ 00:00:40.905 Uttam Kumaran: Yo.

2 00:00:42.850 ⇒ 00:00:43.760 brian: Hey?

3 00:00:46.360 ⇒ 00:00:47.599 brian: Not much. How are you?

4 00:00:47.900 ⇒ 00:00:54.949 Uttam Kumaran: Good man, busy, but good. Yeah. Things are looking up finally, after like months

5 00:00:56.860 ⇒ 00:00:59.695 Uttam Kumaran: on some like things, not looking up.

6 00:01:00.566 ⇒ 00:01:03.830 brian: Why sign some more deals or something?

7 00:01:04.968 ⇒ 00:01:13.940 Uttam Kumaran: Yes, we just signed like 3 new clients. All for like 15 KA month up.

8 00:01:15.240 ⇒ 00:01:16.190 brian: Oh, shit.

9 00:01:16.190 ⇒ 00:01:17.220 brian: Yeah. It’s awesome.

10 00:01:17.360 ⇒ 00:01:18.870 Uttam Kumaran: Yeah. So

11 00:01:19.820 ⇒ 00:01:24.539 Uttam Kumaran: we’re just starting to do a lot more work. And we have like a lot little bit of a bigger team and

12 00:01:24.700 ⇒ 00:01:29.759 Uttam Kumaran: got a little bit better at selling, and so yeah.

13 00:01:31.930 ⇒ 00:01:32.780 brian: That’s great.

14 00:01:32.920 ⇒ 00:01:34.130 brian: I like your new setup.

15 00:01:34.910 ⇒ 00:01:40.510 Uttam Kumaran: Thank you. Yeah, this is my, I live. I’m in East Austin now, like it’s south of Mueller area. So.

16 00:01:42.380 ⇒ 00:01:45.279 brian: 21 0, 2 am.

17 00:01:46.280 ⇒ 00:01:47.130 brian: Yeah.

18 00:01:48.130 ⇒ 00:01:51.880 Uttam Kumaran: Yeah. So how are you? Where are you right now?

19 00:01:53.630 ⇒ 00:01:55.740 brian: I’m in Gran Canario.

20 00:01:56.980 ⇒ 00:01:58.040 Uttam Kumaran: What the fuck is that.

21 00:01:58.980 ⇒ 00:02:00.490 brian: The Canary Islands.

22 00:02:01.220 ⇒ 00:02:01.680 Uttam Kumaran: Where is this?

23 00:02:01.680 ⇒ 00:02:02.500 brian: So there!

24 00:02:03.180 ⇒ 00:02:06.630 brian: There are some islands west of Africa.

25 00:02:07.560 ⇒ 00:02:10.000 brian: but they’re Spanish, so it’s like Spain.

26 00:02:10.000 ⇒ 00:02:11.009 Uttam Kumaran: Crazy.

27 00:02:11.600 ⇒ 00:02:12.820 Uttam Kumaran: So which one.

28 00:02:13.910 ⇒ 00:02:15.100 brian: During Canaria.

29 00:02:16.640 ⇒ 00:02:17.840 Uttam Kumaran: Is that the middle one.

30 00:02:18.930 ⇒ 00:02:19.820 brian: It’s like the relevant.

31 00:02:19.820 ⇒ 00:02:20.939 Uttam Kumaran: The one on the right.

32 00:02:21.760 ⇒ 00:02:24.460 Uttam Kumaran: Wow! So you took a flight here.

33 00:02:26.440 ⇒ 00:02:30.530 brian: Yeah, yeah, I flew. I flew there from Tanzania.

34 00:02:31.000 ⇒ 00:02:36.959 brian: freaking. Longest journey I’ve ever been on. It was like, I think, 31 h or something.

35 00:02:37.620 ⇒ 00:02:40.579 brian: 31 h journey to get here.

36 00:02:41.360 ⇒ 00:02:43.310 Uttam Kumaran: Wow! Holy shit!

37 00:02:44.640 ⇒ 00:02:50.531 brian: Yeah, it’s pretty sweet. It’s last end of last year was pretty crazy. I

38 00:02:52.000 ⇒ 00:02:55.966 brian: I mean, you saw me like not be online and responsive. Much right? Because.

39 00:02:56.880 ⇒ 00:02:59.780 Uttam Kumaran: Because our place in Cape Town got broken into.

40 00:03:00.294 ⇒ 00:03:02.350 brian: What do they take?

41 00:03:02.880 ⇒ 00:03:07.809 brian: They took like my laptop, my air pods, my new ipad. They just took everything.

42 00:03:08.030 ⇒ 00:03:11.690 Uttam Kumaran: And we’re was it a hostel, or was it Airbnb? Or like? What is it.

43 00:03:11.960 ⇒ 00:03:15.775 brian: It was an airbnb, so we just went out for groceries, and then we just kind of

44 00:03:16.670 ⇒ 00:03:21.539 brian: They just crow barred into the into the apartment, and just like took like all of our stuff.

45 00:03:21.780 ⇒ 00:03:37.290 brian: So my laptop, like Abby’s laptop. So then what we had to do is we had to buy a single laptop to share for like 3 weeks, until my parents got here with our actual laptops. So that’s why, like I wasn’t able to respond, or anything, cause I was like half time.

46 00:03:37.290 ⇒ 00:03:38.959 Uttam Kumaran: Do? How do you guys do work?

47 00:03:39.926 ⇒ 00:03:45.249 brian: Well, it’s we work. So it’s like you don’t do much work. So that’s that’s the only reason that worked.

48 00:03:45.250 ⇒ 00:03:46.730 Uttam Kumaran: What does she do for work?

49 00:03:48.176 ⇒ 00:03:53.970 brian: She she’s like college admissions consultant. So she helps like

50 00:03:54.447 ⇒ 00:03:57.120 brian: she helps like fix up essays and things like that

51 00:03:57.280 ⇒ 00:03:58.670 brian: to to get kids into their way.

52 00:03:58.670 ⇒ 00:04:03.840 Uttam Kumaran: But that’s more like just scheduling meetings like when you, whenever you have time, basically.

53 00:04:04.610 ⇒ 00:04:08.330 brian: Yeah, kind of like a lot of like Async work, too. So

54 00:04:08.820 ⇒ 00:04:13.869 brian: we were. We were able to make it work. But it was. It was tough, felt, felt like really behind on everything.

55 00:04:14.870 ⇒ 00:04:15.940 Uttam Kumaran: Oh, really.

56 00:04:16.640 ⇒ 00:04:17.310 brian: Yeah.

57 00:04:17.550 ⇒ 00:04:21.700 Uttam Kumaran: But did you? Did we work? Have any idea you were like? Was it a we work laptop.

58 00:04:22.760 ⇒ 00:04:25.840 brian: So, so I don’t use my. We work laptop. I use my personal laptop.

59 00:04:26.565 ⇒ 00:04:27.010 Uttam Kumaran: Yeah.

60 00:04:27.680 ⇒ 00:04:29.879 brian: Yeah. So it was my personal laptop.

61 00:04:30.470 ⇒ 00:04:33.240 Uttam Kumaran: But like but you it was locked, and everything right. So.

62 00:04:33.240 ⇒ 00:04:38.300 brian: Yeah, yeah, yep, it’s locked and everything. And then I erased it. So it’s fine.

63 00:04:39.040 ⇒ 00:04:44.090 Uttam Kumaran: Okay, got it. Okay, damn dude that sucks.

64 00:04:44.440 ⇒ 00:04:47.240 Uttam Kumaran: So you always. So you got you got one in us and they shipped it.

65 00:04:48.370 ⇒ 00:04:51.590 brian: So my parents came to Cape Town at the end of the year, so they brought.

66 00:04:51.590 ⇒ 00:04:52.140 Uttam Kumaran: Oh, just.

67 00:04:52.140 ⇒ 00:04:55.269 brian: I bought a new laptop, and they brought it for me to Cape Town.

68 00:04:55.450 ⇒ 00:04:58.220 brian: So then I had that whole debacle, and then

69 00:04:58.470 ⇒ 00:05:06.190 brian: I had 2 weeks with my parents, where, like we, we did like Cape Town, and then, like we did a safari in Tanzania with my parents.

70 00:05:06.190 ⇒ 00:05:06.880 Uttam Kumaran: Okay.

71 00:05:07.040 ⇒ 00:05:13.539 brian: So then I was like offline, too. So I’m finally starting to get like a little bit more normalcy.

72 00:05:14.000 ⇒ 00:05:16.630 Uttam Kumaran: Okay, how’s work.

73 00:05:18.090 ⇒ 00:05:25.429 brian: Works fine. It’s it’s like we were not not doing a whole lot. But it’s it’s fine working on that that like

74 00:05:25.880 ⇒ 00:05:29.090 brian: Chatbot thing that I talked to you about the other day.

75 00:05:29.090 ⇒ 00:05:29.780 Uttam Kumaran: Okay.

76 00:05:30.440 ⇒ 00:05:36.710 brian: They’re leading pretty heavy into, like the AI, the AI stuff, and wanting to like have a chat bot stuff like that. So.

77 00:05:36.710 ⇒ 00:05:39.770 Uttam Kumaran: But that’s for the that’s for the data team. Or who is that for Ian.

78 00:05:41.210 ⇒ 00:05:56.420 brian: It’s like we work doesn’t have like. And remember how you used to have like analysts on, like all the teams on all the verticals. Well, they got they got rid of, like all the analysts, basically. So now you just stuck with like business people that don’t know how to write sequel. So they want a chat bot. So those guys can actually do stuff on data.

79 00:05:59.420 ⇒ 00:06:03.651 Uttam Kumaran: Don’t spend much time doing this. That’s you know it’s not. Gonna this is not gonna work.

80 00:06:04.870 ⇒ 00:06:08.855 Uttam Kumaran: Don’t pitch me. Don’t pitch me on this, you know, it’s not gonna work.

81 00:06:09.448 ⇒ 00:06:12.530 brian: I’m not. I’m not pitching you, that’s what they want.

82 00:06:12.530 ⇒ 00:06:16.140 Uttam Kumaran: I know. But, dude, this is the this is like, it’s like.

83 00:06:16.140 ⇒ 00:06:16.880 brian: The direction.

84 00:06:16.880 ⇒ 00:06:25.808 Uttam Kumaran: This is this is like, when you think about AI for 2 min, you’re like, Oh, shit like. I wonder if we could build AI data analyst. And it’s like, Does you have a big brain idea?

85 00:06:26.544 ⇒ 00:06:29.509 brian: Yeah, pretty much. Well, they like.

86 00:06:29.680 ⇒ 00:06:33.170 Uttam Kumaran: That’s the direction that Yuzy and Steven Roland are taking. The teams.

87 00:06:33.820 ⇒ 00:06:37.670 brian: It’s it’s it’s fine. I I just I just do what they tell me to do. So.

88 00:06:38.740 ⇒ 00:06:46.413 Uttam Kumaran: Steven Rowland Bro. What an absolute waste of time!

89 00:06:47.380 ⇒ 00:06:48.320 Uttam Kumaran: Oh, oh, my!

90 00:06:48.320 ⇒ 00:06:48.990 Uttam Kumaran: Oh, yeah!

91 00:06:49.340 ⇒ 00:06:53.170 brian: So I’m working like not a whole lot for them, and I don’t feel bad about it.

92 00:06:53.170 ⇒ 00:06:53.760 Uttam Kumaran: Cortex.

93 00:06:54.930 ⇒ 00:06:58.569 brian: Huh? Yeah, they want to use cortex. Analyst.

94 00:06:58.570 ⇒ 00:07:00.930 Uttam Kumaran: Are you using court? You’re building stuff on cortex.

95 00:07:02.544 ⇒ 00:07:07.899 brian: So we built it, not on cortex. We built it on Gpt initially. And now

96 00:07:08.170 ⇒ 00:07:13.500 brian: and this snowflake came out with cortex. So now we’re gonna move the back end to Cortex Analyst and see how that works.

97 00:07:14.830 ⇒ 00:07:15.570 Uttam Kumaran: Oh, okay.

98 00:07:18.700 ⇒ 00:07:21.449 brian: How about Brainforge? How about how about the projects you guys are working on.

99 00:07:21.450 ⇒ 00:07:27.889 Uttam Kumaran: Things are good dude, I guess. So. Yeah, I don’t know. When is the last time we talked, I guess. Give me a frame of reference, because for me.

100 00:07:28.230 ⇒ 00:07:29.163 Uttam Kumaran: time is

101 00:07:29.630 ⇒ 00:07:30.189 brian: Thank you.

102 00:07:30.390 ⇒ 00:07:34.020 Uttam Kumaran: I am not. My time is like not something I keep like. Have any.

103 00:07:34.020 ⇒ 00:07:34.620 brian: The last

104 00:07:34.900 ⇒ 00:07:40.380 brian: last time I talked I was in Bangkok, and you were pitching me on the tier of price.

105 00:07:41.070 ⇒ 00:07:41.730 Uttam Kumaran: Like, you know.

106 00:07:41.730 ⇒ 00:07:42.839 brian: Overall. Okay? So that.

107 00:07:42.840 ⇒ 00:07:46.260 Uttam Kumaran: Fuck that that’s fuck that here, I’ll show you a couple of things.

108 00:07:48.680 ⇒ 00:07:52.499 Uttam Kumaran: So one is, we basically like have

109 00:07:53.310 ⇒ 00:08:02.289 Uttam Kumaran: several clients that we’re in progress with. We have a couple of Ecom clients.

110 00:08:04.730 ⇒ 00:08:07.920 Uttam Kumaran: One sort of like telehealth. Ecom client?

111 00:08:08.670 ⇒ 00:08:11.860 Uttam Kumaran: We just signed. Have you heard of bolt, bolt dot new.

112 00:08:13.080 ⇒ 00:08:15.600 brian: They’re like they’re like an AI.

113 00:08:15.710 ⇒ 00:08:22.760 Uttam Kumaran: Yeah, they’re like, they’re basically like a almost like Vercel, like a full that’s called Stack Blitz. They’re basically like a full stack

114 00:08:23.620 ⇒ 00:08:26.730 Uttam Kumaran: like quick deployment solution.

115 00:08:27.760 ⇒ 00:08:28.899 brian: For what web apps.

116 00:08:28.900 ⇒ 00:08:32.180 Uttam Kumaran: For web apps and front end and basically front end environments.

117 00:08:32.330 ⇒ 00:08:33.630 Uttam Kumaran: I’ll send it to you on.

118 00:08:35.539 ⇒ 00:08:40.439 Uttam Kumaran: So we’re we just signed them. My friend is basically leading data there.

119 00:08:40.909 ⇒ 00:08:45.219 Uttam Kumaran: And that’s like a pure to come in and do just rip Dvt as fast as possible. Job.

120 00:08:48.150 ⇒ 00:08:48.900 brian: Nice.

121 00:08:52.440 ⇒ 00:08:56.079 Uttam Kumaran: We’re all we’re potentially gonna sign Laravel.

122 00:08:58.150 ⇒ 00:09:00.310 Uttam Kumaran: Look at you on Khp.

123 00:09:00.930 ⇒ 00:09:02.810 brian: Oh, sick! That’s sick!

124 00:09:03.200 ⇒ 00:09:07.390 Uttam Kumaran: Yeah, my a friend that is head of

125 00:09:08.240 ⇒ 00:09:10.540 Uttam Kumaran: like, go to market something there.

126 00:09:11.150 ⇒ 00:09:12.799 Uttam Kumaran: And they have, like no data.

127 00:09:14.500 ⇒ 00:09:16.470 brian: Day. How do these, like massive companies, have no data.

128 00:09:16.782 ⇒ 00:09:23.649 Uttam Kumaran: Was open. Php was open source. They just in the past few years decided to raise money to create a cloud offering.

129 00:09:25.720 ⇒ 00:09:28.370 Uttam Kumaran: Oh, okay, you don’t need data to like.

130 00:09:28.690 ⇒ 00:09:31.249 Uttam Kumaran: Now that I’m on this side, you really can like

131 00:09:31.770 ⇒ 00:09:34.689 Uttam Kumaran: you could go pretty far and not have any idea like.

132 00:09:34.870 ⇒ 00:09:36.460 brian: How it should work.

133 00:09:37.130 ⇒ 00:09:41.799 Uttam Kumaran: And then sometime at some point, the bell like rings, you know, like.

134 00:09:42.792 ⇒ 00:09:44.560 brian: So them, and then.

135 00:09:44.850 ⇒ 00:09:51.669 Uttam Kumaran: We’re also we have a several like AI clients that are sort of at the finish line that we’re sending proposal to. For like building AI agents.

136 00:09:52.350 ⇒ 00:10:01.939 Uttam Kumaran: So we built some agents ourselves internally, like to help with sales and to help with like project management. And then we’re turning that around and sort of selling

137 00:10:03.470 ⇒ 00:10:05.020 Uttam Kumaran: and those are.

138 00:10:05.020 ⇒ 00:10:07.230 brian: What is an AI agent like a Chatbot that.

139 00:10:07.230 ⇒ 00:10:09.300 Uttam Kumaran: It’s no, it’s almost like a

140 00:10:09.430 ⇒ 00:10:14.990 Uttam Kumaran: yeah, I guess. Think less about the medium. It’s more like you tell it to do something. It takes several steps

141 00:10:15.100 ⇒ 00:10:18.940 Uttam Kumaran: on your behalf like it could call other applications workflows.

142 00:10:19.080 ⇒ 00:10:25.520 Uttam Kumaran: It’s basically like an abstraction on top of like, think of it as just building like AI workflows.

143 00:10:26.910 ⇒ 00:10:33.880 Uttam Kumaran: So, for example, for example, we have like a lead research, like, I’ll show you an example. Here’s something that we’re doing. We have like internally like

144 00:10:34.020 ⇒ 00:10:37.240 Uttam Kumaran: literally, we just got sent by now. So Connor, who

145 00:10:37.420 ⇒ 00:10:45.050 Uttam Kumaran: also we just brought him on for sales. He lives here in Austin. He’s a friend of mine. He’s now like on our sales team.

146 00:10:45.509 ⇒ 00:10:57.030 Uttam Kumaran: He, we have a lead research agent that basically you can put in a URL or a Linkedin, and it’ll go basically research everything about the company people and give us suggestions on

147 00:10:59.530 ⇒ 00:11:06.580 Uttam Kumaran: like how to reach out to this client. So this is all an agent that works that an AI workflow that that’s in our slack.

148 00:11:06.910 ⇒ 00:11:08.430 Uttam Kumaran: We also have, like.

149 00:11:08.960 ⇒ 00:11:17.390 Uttam Kumaran: We also all of our meetings, like on Zoom, get recorded. And so they automatically get summarized, classified, and sent to like the right

150 00:11:17.520 ⇒ 00:11:19.850 Uttam Kumaran: sent to the right team as like a summary

151 00:11:21.410 ⇒ 00:11:21.950 brian: Oh, cool!

152 00:11:21.950 ⇒ 00:11:23.720 Uttam Kumaran: That’s 1 thing we’re doing.

153 00:11:24.162 ⇒ 00:11:40.519 Uttam Kumaran: So we have several kind of clients that are like in the proposal stage. There. We also like, have. I have a part time, person that all she does is do notion shit. So we’ve like optimized. Our entire like notion. Workspace is like very clean and like everything is running basically through notion or slack

154 00:11:40.820 ⇒ 00:11:47.700 Uttam Kumaran: or zoom. Now, we don’t really use docs or anything else which is really really nice.

155 00:11:48.110 ⇒ 00:11:54.210 Uttam Kumaran: And then, in terms of data side, the data side is still snowflake or bigquery Dvt.

156 00:11:54.790 ⇒ 00:11:59.700 Uttam Kumaran: and then I’m basically trying to ditch 5 trend, because

157 00:11:59.990 ⇒ 00:12:05.030 Uttam Kumaran: I think their pricing sucks, and I think they starting to suck, so fuck them

158 00:12:05.819 ⇒ 00:12:12.059 Uttam Kumaran: and I want to try like a few other providers like we tried portable dlt hub

159 00:12:14.080 ⇒ 00:12:17.049 Uttam Kumaran: And so we’re trying to make a decision on some better info. There.

160 00:12:17.510 ⇒ 00:12:29.639 Uttam Kumaran: One thing that has one thing that has been helping a lot is like we have scripts for snowflake setup. We have sort of guidelines and naming conventions. I’m working on like linting and things like that on all our repos.

161 00:12:29.760 ⇒ 00:12:33.290 Uttam Kumaran: So getting up to speed on a client is really easy. Now.

162 00:12:35.260 ⇒ 00:12:38.551 Uttam Kumaran: I think our bottleneck one

163 00:12:39.980 ⇒ 00:12:50.570 Uttam Kumaran: We have good people that can take on like Dbt work analyst work. But we don’t have people that are like at my level who can, who are kind of like almost like account managers, or like technical account managers.

164 00:12:50.680 ⇒ 00:12:59.250 Uttam Kumaran: Who can I see the whole project, know the people, and kind of like schmooze, or make sure shit happens like we have project managers.

165 00:12:59.520 ⇒ 00:13:03.120 Uttam Kumaran: And honestly, probably those guys should be doing this. But

166 00:13:03.410 ⇒ 00:13:09.269 Uttam Kumaran: the project managers are sort of just like making sure tickets are like we have one project manager. He’s just making sure tickets are all there and like

167 00:13:09.460 ⇒ 00:13:10.779 Uttam Kumaran: tasks get moved.

168 00:13:13.950 ⇒ 00:13:15.140 Uttam Kumaran: So

169 00:13:16.610 ⇒ 00:13:25.060 Uttam Kumaran: yeah, I think one of the things that we’re working on is, basically, how do? How do we effectively generate a pod for every single new client that we take on?

170 00:13:27.820 ⇒ 00:13:31.820 Uttam Kumaran: And yeah, so at the moment, in terms of like clients.

171 00:13:32.210 ⇒ 00:13:38.580 Uttam Kumaran: Basically, our pitch went from like kind of like booking on an hourly rate. To now we’re doing things more on like a package pricing.

172 00:13:38.850 ⇒ 00:13:46.829 Uttam Kumaran: So we do. We’re trying to do a lot less like hourly, just because it kind of puts us in a bottleneck sometimes that people need more or less work

173 00:13:47.493 ⇒ 00:13:52.870 Uttam Kumaran: and usually like, we’re trying not to go after people who want like less than 10 HA week.

174 00:13:53.620 ⇒ 00:13:57.120 Uttam Kumaran: Because it’s just not enough time to get everything done.

175 00:14:00.170 ⇒ 00:14:07.200 Uttam Kumaran: so I think that’s 1 thing. So we basically for every client we typically are doing like a 5.

176 00:14:07.380 ⇒ 00:14:19.039 Uttam Kumaran: We do like a 1 to 2 week audit where we come in. And for 5 K, we basically will be like, show us all your systems. Everybody involved all your data problems. And we’ll basically give you like a roadmap

177 00:14:19.650 ⇒ 00:14:23.419 Uttam Kumaran: on like, if you need info changes. If you need

178 00:14:23.530 ⇒ 00:14:27.520 Uttam Kumaran: like, sort of strategy changes if you need new infra

179 00:14:27.920 ⇒ 00:14:31.689 Uttam Kumaran: and like, basically break down the entire project. And then we’re like.

180 00:14:31.830 ⇒ 00:14:35.950 Uttam Kumaran: here’s what we would. Here’s how we would. Here’s what we would charge to do that.

181 00:14:36.560 ⇒ 00:14:39.280 Uttam Kumaran: For the most part everybody ends up just going with us.

182 00:14:40.136 ⇒ 00:14:41.229 Uttam Kumaran: You know.

183 00:14:42.200 ⇒ 00:14:42.960 brian: Nice.

184 00:14:44.240 ⇒ 00:14:49.760 brian: So you’re kind of like charging by project, and then you can kind of reuse parts, too. So that’s that’s a good approach.

185 00:14:53.190 ⇒ 00:14:53.615 brian: Sweet.

186 00:14:56.440 ⇒ 00:15:00.029 Uttam Kumaran: So how’s stuff with you like? What do you? What do you think? Hearing that.

187 00:15:03.648 ⇒ 00:15:08.869 brian: No, I think it’s a great idea. You’re like building kind of internal tools, and

188 00:15:09.800 ⇒ 00:15:15.910 brian: can kind of reuse some parts and selling kind of reusable parts that you

189 00:15:16.010 ⇒ 00:15:20.699 brian: built, probably modifying them a little bit right to to clients. Yeah, I think that’s it’s a great approach.

190 00:15:21.740 ⇒ 00:15:26.920 Uttam Kumaran: Yeah, it’s tough, like, I think we’re trying to. I’m trying to lower the cost of goods on every client.

191 00:15:28.450 ⇒ 00:15:33.770 Uttam Kumaran: And honestly, more of the Alpha is on the strategy side, like

192 00:15:34.330 ⇒ 00:15:41.570 Uttam Kumaran: ripping models, ripping Dbt models and stuff. Dude is like whatever. But there’s less people on

193 00:15:41.870 ⇒ 00:15:44.929 Uttam Kumaran: planet Earth that can do this larger strategy.

194 00:15:45.680 ⇒ 00:15:49.860 Uttam Kumaran: like almost like head of data sort of stuff.

195 00:15:50.040 ⇒ 00:15:53.170 Uttam Kumaran: But the nice thing about us is like we we don’t do both

196 00:15:53.330 ⇒ 00:15:56.940 Uttam Kumaran: because there’s a lot of people will be like, I’ll be your head of data. But I won’t do the implementation.

197 00:15:57.210 ⇒ 00:16:00.610 Uttam Kumaran: And there’s a lot of people that like I’ll implement anything. But then you talk to them. They’re like

198 00:16:01.510 ⇒ 00:16:14.299 Uttam Kumaran: Pakistani like you can’t fuck. There’s no video like, you know. It’s like fucking some outsource it firm. There’s not many people that do like full service, and, in fact, a lot of clients. They like us because of that, because we can. I can talk to the CEO. And then also

199 00:16:14.800 ⇒ 00:16:18.109 Uttam Kumaran: basically be like, cool here, we’re gonna rip. Dbt, we’re gonna do. Xyz.

200 00:16:18.750 ⇒ 00:16:23.630 Uttam Kumaran: I think a lot of the things that I’m thinking about on the platform side is

201 00:16:25.380 ⇒ 00:16:28.110 Uttam Kumaran: How do I speed up and improve the quality of

202 00:16:28.330 ⇒ 00:16:32.610 Uttam Kumaran: the requirements that I give to the engineers? And also, how do I speed up?

203 00:16:33.170 ⇒ 00:16:34.400 Uttam Kumaran: How fast they work?

204 00:16:35.070 ⇒ 00:16:41.840 Uttam Kumaran: So I think on the AI side, we’re gonna try. I I don’t know. I think maybe like large, long term, probably like

205 00:16:42.390 ⇒ 00:16:48.759 Uttam Kumaran: a semantic layer or something, is probably best for us to start to automate some of the data tickets that we get

206 00:16:49.050 ⇒ 00:16:50.600 Uttam Kumaran: writing models.

207 00:16:50.940 ⇒ 00:16:58.660 Uttam Kumaran: making column changes things like that. I care very little about automating data analysis, because, frankly, most of that is subjective.

208 00:16:59.501 ⇒ 00:17:08.810 Uttam Kumaran: I care more about automating data modeling like sequel changes new models and automating

209 00:17:09.310 ⇒ 00:17:12.629 Uttam Kumaran: anything on the de side, writing new pipelines, things like that.

210 00:17:12.829 ⇒ 00:17:13.849 Uttam Kumaran: So

211 00:17:14.349 ⇒ 00:17:20.239 Uttam Kumaran: that’s where I think, like there’s real Alpha, because most of that is not client facing. And it’s all code.

212 00:17:22.089 ⇒ 00:17:29.949 brian: Yeah, I mean, is that where mostly the work is coming in right now, like, Oh, I want a new column on my analytics table. Or I need to change this data type.

213 00:17:30.250 ⇒ 00:17:32.549 Uttam Kumaran: It’s kind of coming in. Yeah.

214 00:17:32.890 ⇒ 00:17:37.930 Uttam Kumaran: The Dbt work is so easy because most of these clients were coming. We’re bringing in

215 00:17:40.450 ⇒ 00:17:45.340 Uttam Kumaran: they’re either implementing it for the 1st time, or it’s something fucked up.

216 00:17:45.520 ⇒ 00:17:47.120 Uttam Kumaran: and they’re bringing us in to fix it

217 00:17:47.660 ⇒ 00:17:51.200 Uttam Kumaran: so. But also none. None of these are like

218 00:17:51.680 ⇒ 00:17:55.690 Uttam Kumaran: clients have been using Dvt like, wework’s been using Dvt since 2019

219 00:17:57.100 ⇒ 00:18:01.100 Uttam Kumaran: or late 20.th There’s dude. There’s not many companies that have done that

220 00:18:01.530 ⇒ 00:18:09.209 Uttam Kumaran: right. Actually, there’s people are just finding out about Dbt for the 1st time, like every day.

221 00:18:09.350 ⇒ 00:18:12.859 Uttam Kumaran: So there’s still a lot of people are just going 0 to one.

222 00:18:13.110 ⇒ 00:18:16.540 Uttam Kumaran: So the the actual level of effort is really low. For example.

223 00:18:16.770 ⇒ 00:18:20.280 Uttam Kumaran: every e-commerce company needs the same core entities.

224 00:18:20.800 ⇒ 00:18:24.250 Uttam Kumaran: They need orders, customers, products, shipments

225 00:18:25.330 ⇒ 00:18:29.590 Uttam Kumaran: like all that shit, and it’s the same fucking thing. Every single client.

226 00:18:29.940 ⇒ 00:18:36.619 Uttam Kumaran: What’s different, though, is the client is different, and what they care about the business model may be different.

227 00:18:36.920 ⇒ 00:18:52.579 Uttam Kumaran: But that doesn’t mean the core data entities are any different. Right? What’s different is the sources. Sometimes people use different shipment providers, things like that. But the core data models typically stay the same. So you’re fitting whatever data you get into those core models. This is why I think the activity. Schema

228 00:18:53.240 ⇒ 00:18:58.630 Uttam Kumaran: is probably more long term, like more interesting to me.

229 00:19:02.590 ⇒ 00:19:04.919 Uttam Kumaran: In terms of just like creating sort of

230 00:19:05.300 ⇒ 00:19:10.610 Uttam Kumaran: one place where you can shove everything in and then building these core entities of.

231 00:19:11.710 ⇒ 00:19:14.049 Uttam Kumaran: But also I don’t know I feel like, still, it’s

232 00:19:14.380 ⇒ 00:19:17.340 Uttam Kumaran: not. There’s not a lot of great Dbt people so

233 00:19:17.610 ⇒ 00:19:24.260 Uttam Kumaran: to explain like having to teach people how to do activity. Schema, which is kind of like a Meta concept, is like.

234 00:19:24.700 ⇒ 00:19:28.730 Uttam Kumaran: I don’t know. I don’t. Wanna I don’t wanna do that. So yeah.

235 00:19:31.040 ⇒ 00:19:37.669 brian: Yeah, so, yeah, I guess. What are you asking me like, what like, I.

236 00:19:37.670 ⇒ 00:19:43.219 Uttam Kumaran: I just. I’m just curious on like when you hear that like, what do you? What do you think like?

237 00:19:43.760 ⇒ 00:19:47.360 Uttam Kumaran: I don’t know. We’re not a lot of the problems are solving. The Dbt side are like

238 00:19:48.160 ⇒ 00:19:53.779 Uttam Kumaran: just like basics. Still. So for me, what’s interesting is like building

239 00:19:54.020 ⇒ 00:19:57.969 Uttam Kumaran: a good library of internal Macros building good Cicd.

240 00:19:58.140 ⇒ 00:20:03.400 Uttam Kumaran: building, a good building, a good basically building quality of life stuff for our developers

241 00:20:04.090 ⇒ 00:20:08.049 Uttam Kumaran: to push less shitty code and to write code faster.

242 00:20:08.450 ⇒ 00:20:12.880 Uttam Kumaran: The second thing I’m is like, how do I get AI to write fucking? Dbt code?

243 00:20:14.950 ⇒ 00:20:19.829 brian: Yeah. So I haven’t done a lot of like ae work, right? So. But it sounds like.

244 00:20:19.830 ⇒ 00:20:26.330 Uttam Kumaran: The the amount of work that you’ve done is is way more than the stuff we’re doing. I’m telling you like

245 00:20:26.610 ⇒ 00:20:30.169 Uttam Kumaran: you’re familiar with. You know how to write like Dvt. Models.

246 00:20:31.250 ⇒ 00:20:32.040 brian: Yeah.

247 00:20:33.000 ⇒ 00:20:34.180 Uttam Kumaran: That’s all. That’s all. It is

248 00:20:34.590 ⇒ 00:20:36.780 Uttam Kumaran: fucking select statement. You’re writing a query, and you’re

249 00:20:37.240 ⇒ 00:20:42.540 Uttam Kumaran: shoving it into there. That’s it, that’s all we’re doing. We’re not. We’re not doing any like

250 00:20:42.850 ⇒ 00:20:51.640 Uttam Kumaran: we’re. We’re not even getting to the point where we like have tons of different jobs like. And we have like different environments. And all that stuff like these are not that

251 00:20:52.580 ⇒ 00:20:53.199 Uttam Kumaran: they’re not.

252 00:20:53.200 ⇒ 00:20:54.110 brian: Yeah, so like.

253 00:20:54.410 ⇒ 00:21:07.980 brian: So I think what you need is like, just like a template right like, Oh, you sign another e-commerce person. Then just like this is the starting template for e-commerce. Not just tweak like tweak it a little bit so that it fits for this new e-commerce one.

254 00:21:08.490 ⇒ 00:21:11.310 Uttam Kumaran: I think that the challenge is that their sources are different.

255 00:21:13.360 ⇒ 00:21:20.500 brian: Then you abstract the the sources away into like variables where you set them right, like I I don’t know how reusable each

256 00:21:20.630 ⇒ 00:21:23.480 brian: e-commerce dbt thing is.

257 00:21:25.050 ⇒ 00:21:28.090 brian: or or you could build it in like layers. I guess I’m just kind of thinking out loud.

258 00:21:28.090 ⇒ 00:21:30.090 Uttam Kumaran: No, it’s like it’s like the Dbt like.

259 00:21:30.200 ⇒ 00:21:41.039 Uttam Kumaran: The intermediate layer is probably reusable, and the March layer is probably reusable. The raw layer, like how we fit the sources into. That is the challenge.

260 00:21:41.160 ⇒ 00:21:57.929 Uttam Kumaran: because we have clients who like have some of the data, or they have it in other different formats, or they’re like they have different, like, for example, they have different granularities that maybe we have to account for. So there’s some sort of business stuff that’s not like, extremely clean.

261 00:21:59.310 ⇒ 00:22:00.120 brian: So

262 00:22:00.370 ⇒ 00:22:11.180 brian: if the work is in the raw layer kind of like what bronze, silver, gold, whatever right? And the work is in the bronze layer. Then you make templates for the silver and the gold layer.

263 00:22:11.420 ⇒ 00:22:13.219 brian: and then on the bronze layer.

264 00:22:14.220 ⇒ 00:22:17.589 Uttam Kumaran: The bronze layer is like I need to. I need to fit

265 00:22:18.440 ⇒ 00:22:21.539 Uttam Kumaran: one or many sources into that like sort of mold.

266 00:22:22.390 ⇒ 00:22:33.390 brian: Yeah, that that to me just sounds like your, your, your like brain forge work. Right? Someone has to just do that, because that seems annoying to automate like you can make probably some AI stuff to

267 00:22:33.720 ⇒ 00:22:40.720 brian: get you, maybe like 50% of the way there. But it sounds like that part is like where the actual hours are.

268 00:22:40.960 ⇒ 00:23:04.600 Uttam Kumaran: Yeah. Because here’s here’s how I feel is like even the Dbt work like, let’s say you want to write an orders table. At this point. I’m like dude. I told my I told people on our team, said, shove all of the constituent source tables, shove the Ddls and the table you want into chat. Gpt, have it right, the fucking model. Do not do not write that query from scratch

269 00:23:04.910 ⇒ 00:23:06.760 Uttam Kumaran: like you’re wasting time.

270 00:23:09.080 ⇒ 00:23:11.710 Uttam Kumaran: Explain what the output query you want.

271 00:23:11.890 ⇒ 00:23:15.809 Uttam Kumaran: shove the data, and maybe snippets of what the shape looks like.

272 00:23:16.360 ⇒ 00:23:19.689 Uttam Kumaran: And then you should be able to just write those models right? Like, what’s the

273 00:23:20.790 ⇒ 00:23:23.110 Uttam Kumaran: why are we still writing like?

274 00:23:23.770 ⇒ 00:23:25.690 Uttam Kumaran: And also for making updates

275 00:23:26.040 ⇒ 00:23:30.769 Uttam Kumaran: right? Like, why can’t an AI this, I’m sure I think of it like, why can’t you tell an AI

276 00:23:31.110 ⇒ 00:23:37.410 Uttam Kumaran: in the orders table? There is a customer segment case. When I want to add

277 00:23:37.860 ⇒ 00:23:48.380 Uttam Kumaran: an another clause that does this, the AI should be able to go isolate that file, make the change, commit it, and open the Pr.

278 00:23:54.200 ⇒ 00:24:02.089 brian: Well, the stuff later is just like automating like processes. That stuff is easy, like what the tricky part is is the AI getting

279 00:24:02.430 ⇒ 00:24:04.440 brian: getting the SQL. Query correct.

280 00:24:04.560 ⇒ 00:24:07.250 brian: That’s the that’s the part tricky part right?

281 00:24:09.210 ⇒ 00:24:12.959 Uttam Kumaran: But I think it’s more of like a requirements problem like

282 00:24:13.400 ⇒ 00:24:16.710 Uttam Kumaran: we just need to constrain the types of things it can do.

283 00:24:16.880 ⇒ 00:24:19.649 Uttam Kumaran: So you start small column, Renames.

284 00:24:20.390 ⇒ 00:24:23.710 Uttam Kumaran: then you move on to something else. Then you move on to something else. Right?

285 00:24:29.820 ⇒ 00:24:33.730 Uttam Kumaran: Because, dude, that is what I’m saying. I don’t. I don’t care about solving the analysis piece.

286 00:24:34.230 ⇒ 00:24:35.750 Uttam Kumaran: It’s not solvable.

287 00:24:35.880 ⇒ 00:24:37.330 Uttam Kumaran: I’m telling you.

288 00:24:37.510 ⇒ 00:24:46.359 Uttam Kumaran: I talked to some people who try who are doing this same business. They’re trying to automate data analysts. They’ve been doing it now for 2 years. They’re not going anywhere

289 00:24:47.860 ⇒ 00:24:59.029 Uttam Kumaran: like it’s not gonna happen that it’s not happening. Customers are not asking like, what are my orders today? Nobody is asking that fucking, dumb ass question. They’re asking hard

290 00:24:59.260 ⇒ 00:25:23.489 Uttam Kumaran: questions where you need to like, really have? You need to have, like some sort of like understanding of the business model. You need to go talk to people. Then you need to look at the data. It’s it’s like, maybe AI will solve in 5 years. But that’s not. I’m not looking to for the AI to to come up with a query for analysis. I’m looking for cement. I’m looking for change stuff in Dbt.

291 00:25:24.370 ⇒ 00:25:29.989 Uttam Kumaran: and I’m looking for like help with basically executing stuff on Snowflake.

292 00:25:32.660 ⇒ 00:25:36.299 brian: Yeah, okay, so those are like 2 different problems like the Dbc problem.

293 00:25:36.820 ⇒ 00:25:37.989 brian: That just depends how good.

294 00:25:37.990 ⇒ 00:25:38.590 Uttam Kumaran: Yeah, I didn’t.

295 00:25:38.590 ⇒ 00:25:42.439 brian: Depends how good your model is. Right, like the how good the Gpt model is like.

296 00:25:42.560 ⇒ 00:25:58.189 brian: like, there’s a reason like, I don’t know. Google hasn’t been able to say, Change this button to red right? And they just have it all be automated by AI. And that’s kind of what you’re trying to do with like the data bottles. And I think it’s just like Gbc, 4, maybe is just is not good enough, or the prompting is.

297 00:25:58.475 ⇒ 00:26:05.050 Uttam Kumaran: Think it is good enough. I think it’s more about like you need to get the right context in at the right time.

298 00:26:05.530 ⇒ 00:26:15.349 Uttam Kumaran: right? Like dude if you take if you take a I mean dude. The problem thing is, I code. I I use copilot with Dbt coding every day. It’s totally good.

299 00:26:15.790 ⇒ 00:26:22.010 Uttam Kumaran: I’m trying to take it one step further. I don’t want it to do auto, but I want you to be able to be like, go make this change

300 00:26:22.380 ⇒ 00:26:29.730 Uttam Kumaran: go like, for example, we, one of the things we literally just spent hours doing yesterday is literally

301 00:26:29.870 ⇒ 00:26:34.640 Uttam Kumaran: go make an orders table that pulls 3 columns from here, and 3 columns from here.

302 00:26:37.540 ⇒ 00:26:40.860 Uttam Kumaran: The AI. Should be able to go solve that and propose the Pr.

303 00:26:42.410 ⇒ 00:26:49.909 Uttam Kumaran: so that’s 1 thing. So that’s that’s 1 thing. The second flip version of this is, I want to use AI in the Pr review process.

304 00:26:50.880 ⇒ 00:26:58.229 Uttam Kumaran: It’s basically like linting on like steroids like, I want the AI to understand what the ask was.

305 00:26:58.410 ⇒ 00:27:00.920 Uttam Kumaran: Look at the Pr and basically say, like.

306 00:27:01.180 ⇒ 00:27:02.919 Uttam Kumaran: this was good enough or bad.

307 00:27:03.880 ⇒ 00:27:07.810 Uttam Kumaran: right? Because one of the big ways to ensure code quality is code reviews.

308 00:27:08.040 ⇒ 00:27:15.219 Uttam Kumaran: But one of the problems with code reviews is that it involves more than one person, which means the shit takes longer.

309 00:27:15.590 ⇒ 00:27:17.690 Uttam Kumaran: It like slows. It like

310 00:27:17.950 ⇒ 00:27:26.070 Uttam Kumaran: stuff stays in review for a fucking day when it’s basically already done. And then when you review it, you’re like you missed this thing, and then they take another day to

311 00:27:26.450 ⇒ 00:27:30.870 Uttam Kumaran: that. Stuff is all wasted time for me. I’m like, when you submit a Pr.

312 00:27:30.970 ⇒ 00:27:40.789 Uttam Kumaran: we, we basically want to build an AI Review bot that looks at the code has full understanding of Dbt, and then can actually go say, like, this is a good Pr or not.

313 00:27:46.320 ⇒ 00:27:52.880 brian: I think that’s a harder problem to solve than like, add a column or make a new orders table using these 3 and these 3. I think that’s a harder problem.

314 00:27:52.880 ⇒ 00:27:58.190 Uttam Kumaran: Really, I think it’s the opposite. I think what I think. I think what I said 1st is a harder problem.

315 00:27:58.710 ⇒ 00:28:01.419 brian: Because what’s I think? What’s hard about? The second problem

316 00:28:01.800 ⇒ 00:28:04.850 brian: is, how do you know it’s a good like

317 00:28:04.950 ⇒ 00:28:09.039 brian: a Pr comes Tom. How how do you know it’s a good Pr.

318 00:28:10.210 ⇒ 00:28:15.960 brian: you have that domain knowledge of what this thing is right. And if you don’t have Domain Lodge, you gotta like read through.

319 00:28:16.210 ⇒ 00:28:20.690 Uttam Kumaran: All of these like sequel files. To understand what you’re doing is just like, yeah.

320 00:28:20.690 ⇒ 00:28:21.610 brian: This legit right.

321 00:28:21.610 ⇒ 00:28:23.080 Uttam Kumaran: Read through all the files.

322 00:28:23.690 ⇒ 00:28:29.215 brian: I think that’s a harder. That’s like a bigger context. For, like an Llm. To have to.

323 00:28:29.540 ⇒ 00:28:31.709 Uttam Kumaran: Knows everything about the business

324 00:28:32.340 ⇒ 00:28:40.650 Uttam Kumaran: like like, let’s say we provide it. Here’s the client. Here’s what we’re tasked to do for them. Here’s all the Snowflake docs. Here’s all the Dbt Docs

325 00:28:41.470 ⇒ 00:28:47.969 Uttam Kumaran: 1st classify what type of Pr, this is. Second, look at the Pr tech ticket which has notes.

326 00:28:49.350 ⇒ 00:28:54.660 Uttam Kumaran: And then judge this basically or suggest changes.

327 00:28:58.520 ⇒ 00:29:12.600 Uttam Kumaran: And and honestly, this is the thing it doesn’t even need for the most part, most Prs won’t like. I don’t know whether I don’t know what the ratio is. Prs, with changes versus not, but for the most part it can just say there’s an issue

328 00:29:12.810 ⇒ 00:29:13.929 Uttam Kumaran: and escalate

329 00:29:14.730 ⇒ 00:29:20.230 Uttam Kumaran: that at least cuts down some of the ones that don’t, because I don’t want to sit here reviewing Prs that have no problems.

330 00:29:21.450 ⇒ 00:29:25.740 Uttam Kumaran: But I I still wanna have code owners. I still wanna sort of have a process.

331 00:29:25.740 ⇒ 00:29:30.250 brian: So here’s a question in in your ideal world. Does someone look at?

332 00:29:30.754 ⇒ 00:29:35.649 brian: Does someone look at every single Pr or does some Pr. Just make it through without anyone looking at it.

333 00:29:35.920 ⇒ 00:29:40.689 Uttam Kumaran: You know the you know the answer. It’s like, because someone has got to look at every Pr. But we don’t do that.

334 00:29:40.690 ⇒ 00:29:41.050 brian: Yeah.

335 00:29:41.050 ⇒ 00:29:48.420 Uttam Kumaran: We’re fucking lazy. We’re just like fucking. Send it. This is the thing. AI is not lazy. AI will look at every single thing.

336 00:29:48.610 ⇒ 00:29:55.830 Uttam Kumaran: So that’s the sort of stuff where and to give you to tell you what’s actually happening, I will send you some companies that are doing this

337 00:29:56.220 ⇒ 00:30:05.419 Uttam Kumaran: Github Pr review comment automation. But they’re doing it in in like full stack edge.

338 00:30:05.740 ⇒ 00:30:09.589 Uttam Kumaran: They’re not doing it in data. Nobody’s doing this stuff in data

339 00:30:10.580 ⇒ 00:30:20.770 Uttam Kumaran: like nobody’s building a Dbt Pr review. Bot, they’re trying to solve data analysis

340 00:30:22.560 ⇒ 00:30:25.540 Uttam Kumaran: like dude. It’s so moronic like.

341 00:30:26.930 ⇒ 00:30:34.739 brian: Well, yeah, the reason you can’t solve that is because you need domain knowledge. I think at the end of the day the more domain knowledge you need, the harder it is for the Llm. To solve. That’s why.

342 00:30:35.260 ⇒ 00:30:35.750 Uttam Kumaran: Serving.

343 00:30:35.750 ⇒ 00:30:41.300 Uttam Kumaran: It’s subjective like it matters the presentation matters who you’re presenting to.

344 00:30:42.370 ⇒ 00:30:46.340 Uttam Kumaran: Some people don’t fuck with looker. They’re like, I only want tableau.

345 00:30:47.760 ⇒ 00:30:49.080 brian: Yeah, yeah, it’s it’s like, an.

346 00:30:49.080 ⇒ 00:30:52.070 Uttam Kumaran: You’re not gonna know, or they’re like they’re like.

347 00:30:52.310 ⇒ 00:30:58.280 Uttam Kumaran: or they just don’t have information that they’re gatekeeping. And when you make the presentation, they’re like, Oh, actually, I forgot to tell you this.

348 00:30:59.640 ⇒ 00:31:06.099 Uttam Kumaran: That’s analysis work, dude analysis work is like mostly that it’s not actually like.

349 00:31:06.620 ⇒ 00:31:09.650 Uttam Kumaran: It’s not actually that much a day. I don’t think it’s like.

350 00:31:10.260 ⇒ 00:31:14.940 Uttam Kumaran: I think it’s like 50% people, 50% data work like.

351 00:31:14.940 ⇒ 00:31:17.100 brian: Yeah, it’s mostly it’s just like presenting.

352 00:31:17.100 ⇒ 00:31:42.080 Uttam Kumaran: That’s why we have Looker Sigma power. Bi, we have, we have stream lay, we have, you know, because you know why? Because it’s subjective, otherwise we would have one. We would have dbt, like we would have Snowflake. We have 3 different data warehouse to be all the same features at the Bi level. It’s it’s so subjective. That there’s it’s there’s no competitive advantage. So it’s not worth

353 00:31:43.130 ⇒ 00:31:45.689 Uttam Kumaran: automating those sort of like.

354 00:31:46.180 ⇒ 00:31:59.029 Uttam Kumaran: And again, I work with executives all day. Executives are not data. People think that executives are like, what’s my sales yesterday for? People in California dude? No, they’re asking like

355 00:31:59.220 ⇒ 00:32:12.890 Uttam Kumaran: we? We fucking harder questions they’re asking like, Yo, what’s our like? Cohort retention for every person that goes to this landing page 1st versus this one. Okay, cool. Now, I need amplitude. Okay, cool. Now, I need like

356 00:32:13.040 ⇒ 00:32:16.819 Uttam Kumaran: sales data. Oh, shit our sales. Data is like fuck. They got a back.

357 00:32:16.940 ⇒ 00:32:21.743 Uttam Kumaran: You know, there’s like all these issues, it’s not that easy.

358 00:32:22.660 ⇒ 00:32:23.440 brian: Yeah.

359 00:32:23.440 ⇒ 00:32:24.070 Uttam Kumaran: Yeah.

360 00:32:25.250 ⇒ 00:32:30.439 brian: Okay. So the 2 things that you’re you’re or the it sounds like the main thing you’re trying to work on is the

361 00:32:31.130 ⇒ 00:32:34.159 brian: the dbt, a. Pr. Review, bot.

362 00:32:34.340 ⇒ 00:32:40.530 Uttam Kumaran: I guess this is where I’m more. I’m curious about how we can start to use

363 00:32:41.006 ⇒ 00:32:48.979 Uttam Kumaran: AI in a more focused way on the engineering side, and from everything that I’ve like, I talked to a lot of vendors.

364 00:32:49.270 ⇒ 00:32:51.419 Uttam Kumaran: And I talked to a lot of people about

365 00:32:51.620 ⇒ 00:32:57.109 Uttam Kumaran: automating this stuff. It’s happening in back end engineering and front end. It’s not happening in data.

366 00:32:58.720 ⇒ 00:33:08.060 Uttam Kumaran: Everybody in data is focused on the data analysis problem. Nobody’s focused on automating Dbt work or automating pipeline development.

367 00:33:09.290 ⇒ 00:33:12.259 Uttam Kumaran: So I do think that there’s an opportunity for us to

368 00:33:12.370 ⇒ 00:33:16.180 Uttam Kumaran: use Brainforge as a good guinea pig to assess

369 00:33:16.560 ⇒ 00:33:18.969 Uttam Kumaran: the feasibility of some of those solutions.

370 00:33:20.860 ⇒ 00:33:23.289 Uttam Kumaran: In order to basically for us.

371 00:33:23.610 ⇒ 00:33:25.730 Uttam Kumaran: the kind of the advantage is one

372 00:33:26.190 ⇒ 00:33:32.169 Uttam Kumaran: we can have less skilled people because the AI will supplement in some ways.

373 00:33:32.839 ⇒ 00:33:37.770 Uttam Kumaran: Or we can. It takes. We can train. It takes. We can train people faster. The second thing is

374 00:33:40.500 ⇒ 00:33:43.470 Uttam Kumaran: So the 1st thing is about, yeah, we. So the

375 00:33:43.900 ⇒ 00:33:47.469 Uttam Kumaran: the second thing is mainly what was, I think the second thing about.

376 00:33:48.910 ⇒ 00:33:57.289 Uttam Kumaran: oh, yeah, we just deliver faster. And then, basically a lot of the clients we’re already winning is because we’re doing things way faster. And that’s just

377 00:33:57.440 ⇒ 00:33:59.850 Uttam Kumaran: that’s just because that’s just scale.

378 00:34:00.040 ⇒ 00:34:06.560 Uttam Kumaran: But like, I think the AI can probably add 20 or 30% more in terms of speed.

379 00:34:08.460 ⇒ 00:34:11.490 brian: Okay, so this is, this is kind of how I would do it.

380 00:34:15.020 ⇒ 00:34:17.690 brian: For just like delivering things faster.

381 00:34:17.889 ⇒ 00:34:25.830 brian: So someone writes like product manager writes, detailed Jira ticket about what change they want.

382 00:34:26.170 ⇒ 00:34:29.219 brian: Then the AI would

383 00:34:30.100 ⇒ 00:34:37.870 brian: try to figure out what that is, and this AI is already trained on your notion stuff. So it already has context of the business.

384 00:34:38.020 ⇒ 00:34:42.740 brian: Right? Try to figure out what that is, and then what that’s what’s that

385 00:34:43.150 ⇒ 00:34:51.279 brian: that’s sent to whatever engineer picks up the ticket, and ideally they can just open it up and already see the AI’s proposed changes

386 00:34:51.389 ⇒ 00:34:55.470 brian: and what the ticket description is, and then they can make

387 00:34:56.500 ⇒ 00:35:03.449 brian: they can look at it, see if it’s right or wrong. If it’s wrong, they fix it. If it’s right, then they push, they

388 00:35:03.650 ⇒ 00:35:09.689 brian: submit the Pr. I don’t know what happens after they submit the Pr. I don’t know if that’s like someone else reviews it, or.

389 00:35:10.960 ⇒ 00:35:11.640 Uttam Kumaran: It just gets quick.

390 00:35:11.640 ⇒ 00:35:14.240 brian: It just gets pushed or whatever. But

391 00:35:15.000 ⇒ 00:35:24.059 brian: that’s that’s kind of how I do it. Because that that way you’re kind of like the the minimum things you need to solve. This kind of problem is you need requirements, right? And.

392 00:35:24.060 ⇒ 00:35:24.630 Uttam Kumaran: Dennis.

393 00:35:24.630 ⇒ 00:35:29.290 brian: The someone has to write requirements. So your pro, your Pm. Or whatever.

394 00:35:29.290 ⇒ 00:35:38.860 Uttam Kumaran: Let’s assume the requirements are fire dude, and all this stuff it requires. AI only works. So the requirements are extremely, extremely detailed, so you can assume.

395 00:35:38.860 ⇒ 00:35:43.909 brian: So whoever someone pushes in the requirements onto a jury ticket, you have something. Pick up that jury ticket.

396 00:35:44.300 ⇒ 00:35:45.579 Uttam Kumaran: Gets linked to the Pr.

397 00:35:46.010 ⇒ 00:36:00.360 brian: It gets linked to the Pr. Tries to figure out what what the proposed change is makes the makes the proposed change for Pr. Then the engineer, you, or whoever or me, or whatever checks out that pr see if it’s right or wrong.

398 00:36:00.730 ⇒ 00:36:13.970 brian: or just just like checks the Pr. If it’s if the Pr. Simple like, rename this column, then like it, takes a second for for that engineer, and that they don’t have to hand it off again. They can just either merge it or like. If it’s wrong, they check it out, they fix it, then they merge it, or something like that.

399 00:36:16.170 ⇒ 00:36:17.780 Uttam Kumaran: Yeah, that’s pretty. It.

400 00:36:18.560 ⇒ 00:36:19.400 brian: I think that’s like

401 00:36:19.400 ⇒ 00:36:25.220 brian: my question from for you. Now, since you’re kind of like looking at cortex stuff like.

402 00:36:25.790 ⇒ 00:36:31.759 Uttam Kumaran: Does, being on snowflake like, give us any advantage on any of these sort of like AI things.

403 00:36:32.020 ⇒ 00:36:33.479 Uttam Kumaran: I wonder if it’s like.

404 00:36:34.990 ⇒ 00:36:39.659 Uttam Kumaran: yeah, like. I wonder how we can use cortex for this process like a little bit better.

405 00:36:40.440 ⇒ 00:36:50.020 brian: Snowflake. No, I don’t think Snowflake gives you much advantage. What the problem Snowflake is trying to solve is people don’t know how to write sequel, and people don’t know

406 00:36:50.130 ⇒ 00:36:58.410 brian: how these tables and entities are related to each other. So they want, you know, someone who’s a not even a business analyst, just like.

407 00:36:58.410 ⇒ 00:37:00.629 Uttam Kumaran: That’s like bad document. That’s like

408 00:37:00.630 ⇒ 00:37:02.280 Uttam Kumaran: they just want someone to. They just want that

409 00:37:02.280 ⇒ 00:37:04.499 Uttam Kumaran: dude. That’s not the real problem.

410 00:37:04.890 ⇒ 00:37:11.630 brian: They just want that person who doesn’t know anything about the data to be able to go in and just ask, what is my sales.

411 00:37:11.630 ⇒ 00:37:13.078 Uttam Kumaran: What is my data?

412 00:37:13.440 ⇒ 00:37:22.669 brian: Exactly exactly. Then they they can ask those questions without sending that message into slack. And then, having, like a data analyst, come and like. Figure it out. That’s what cortex is.

413 00:37:23.110 ⇒ 00:37:24.750 brian: in my opinion, trying to solve.

414 00:37:25.260 ⇒ 00:37:25.890 Uttam Kumaran: Okay.

415 00:37:28.370 ⇒ 00:37:28.920 brian: So for

416 00:37:28.920 ⇒ 00:37:36.710 brian: this kind of problem, you you need a Llm that’s trained on your like massive notion. With all of your context, you need. You need something that’s

417 00:37:37.830 ⇒ 00:37:39.730 brian: that has that domain knowledge.

418 00:37:40.160 ⇒ 00:37:41.849 Uttam Kumaran: We can already do that. We already do that.

419 00:37:43.240 ⇒ 00:37:49.929 Uttam Kumaran: So like this is where it’s like us, the AI challenges that you described. We’ve we’ve already. We do this. So we have.

420 00:37:50.120 ⇒ 00:37:56.940 Uttam Kumaran: We already have a notion that’s that already has context, our entire notion. We have a, we have AI agents context of our entire notion.

421 00:37:57.160 ⇒ 00:38:01.480 Uttam Kumaran: We also have built an AI engine that has context of all of snowflake stocks.

422 00:38:01.640 ⇒ 00:38:03.410 Uttam Kumaran: All the Dbt docs.

423 00:38:03.940 ⇒ 00:38:08.680 Uttam Kumaran: Right? So so assume those are solved problems. I think for me, the biggest problem is thinking about

424 00:38:09.110 ⇒ 00:38:13.499 Uttam Kumaran: what are the nuances of this particular like data challenge, like.

425 00:38:13.740 ⇒ 00:38:18.910 Uttam Kumaran: what does the Asian need to be particular about like, what would we check

426 00:38:20.700 ⇒ 00:38:25.259 Uttam Kumaran: like? What sort of review process would we go through? And how do we replicate that?

427 00:38:25.720 ⇒ 00:38:28.439 Uttam Kumaran: And more about like, what is the data side of it like.

428 00:38:30.620 ⇒ 00:38:35.639 Uttam Kumaran: like, I guess it would need to have all files. It would need to know, like, what sort of change you’re trying to make.

429 00:38:36.780 ⇒ 00:38:41.539 brian: Well, if you already have all those agents trained, then you just need to like, put the pieces together right. You need.

430 00:38:41.570 ⇒ 00:38:42.270 Uttam Kumaran: Yes.

431 00:38:42.270 ⇒ 00:38:48.390 brian: Some sort of some sort of hook for your ticketing system, for when a hook, when a ticket comes in, and then that’s it.

432 00:38:48.390 ⇒ 00:38:50.989 Uttam Kumaran: Automatically sent to the reviewer and the reviewer.

433 00:38:51.520 ⇒ 00:38:51.990 brian: And then that.

434 00:38:52.240 ⇒ 00:38:52.990 Uttam Kumaran: Checks it out.

435 00:38:53.660 ⇒ 00:38:55.430 brian: Well, the 1st thing you need is like

436 00:38:55.570 ⇒ 00:38:59.999 brian: the hook into the ticket ticket comes in that picks up the ticket.

437 00:39:00.290 ⇒ 00:39:09.010 brian: writes your Pr. For you, and then submits a Pr. And then and then you just need an engineer to come. Pick up that Pr to see if it’s like any good.

438 00:39:11.290 ⇒ 00:39:15.610 brian: So you, if you already have all the pieces you just need to put them together with some sort of like automation pipeline.

439 00:39:19.140 ⇒ 00:39:23.970 Uttam Kumaran: Yeah, I guess I I guess I’ll need to figure out also like how to do the checkout process like.

440 00:39:24.230 ⇒ 00:39:26.449 Uttam Kumaran: And also I don’t know how the actual like?

441 00:39:28.030 ⇒ 00:39:31.210 Uttam Kumaran: How does it make the change and then commit it.

442 00:39:34.490 ⇒ 00:39:36.930 brian: How does it make the change in commit it?

443 00:39:38.230 ⇒ 00:39:42.980 Uttam Kumaran: Like. For example, let’s say we have a ticket. That’s just as simple as like, go change this column name.

444 00:39:43.260 ⇒ 00:39:45.700 Uttam Kumaran: and I want to say, cool. Give that to AI

445 00:39:46.800 ⇒ 00:39:48.559 Uttam Kumaran: trying to think about like so.

446 00:39:49.110 ⇒ 00:39:53.659 Uttam Kumaran: and assume like our Ci CD process will take care of like the does it? Does it run

447 00:39:58.270 ⇒ 00:40:01.110 brian: You know what I will put on. The ticket type is like

448 00:40:01.761 ⇒ 00:40:04.500 brian: type, like change type, change column type.

449 00:40:04.810 ⇒ 00:40:08.299 brian: add column, change column, name that would probably help the AI a lot.

450 00:40:08.440 ⇒ 00:40:16.520 brian: or if there’s like an other more complex than you know, maybe the AI is not not good at handling it right then you have different processes for it.

451 00:40:19.790 ⇒ 00:40:27.339 Uttam Kumaran: Yeah, okay, yeah, I’m gonna I don’t know. I think. Still, we’re probably like 3 or

452 00:40:28.940 ⇒ 00:40:31.510 Uttam Kumaran: more likely 6 months away from expending

453 00:40:31.710 ⇒ 00:40:35.320 Uttam Kumaran: more time on it, just because it’s like this is not the.

454 00:40:35.700 ⇒ 00:40:37.909 Uttam Kumaran: This is not our number one problem. But

455 00:40:39.290 ⇒ 00:40:44.309 Uttam Kumaran: I will say what our number one problem will be is getting good people like it’s hard to find.

456 00:40:44.950 ⇒ 00:40:49.410 Uttam Kumaran: It’s I mean, it’s hard to find people like me and you for cheap and not even cheap like

457 00:40:50.340 ⇒ 00:40:56.839 Uttam Kumaran: for just like a normal amount of money. And I’m speaking on both sides, like, you know. So

458 00:40:57.000 ⇒ 00:41:00.129 Uttam Kumaran: it’s tough. On one hand, I want to hire more people.

459 00:41:00.490 ⇒ 00:41:05.260 Uttam Kumaran: but the only way I can solve that is, by charging more rates, and

460 00:41:05.480 ⇒ 00:41:09.030 Uttam Kumaran: the only way to increase the margin in this business is to

461 00:41:10.060 ⇒ 00:41:12.369 Uttam Kumaran: sort of augment that a little bit.

462 00:41:12.960 ⇒ 00:41:19.730 Uttam Kumaran: And I think AI is gonna help solve a little bit of that. And then I want me. People like me and you to go work on like, really, the tougher

463 00:41:19.980 ⇒ 00:41:21.160 Uttam Kumaran: problems like

464 00:41:21.720 ⇒ 00:41:27.469 Uttam Kumaran: that’s the thing I want, like an hour of the smartest people to go to like the really, the app, the work that we can build

465 00:41:27.810 ⇒ 00:41:29.539 Uttam Kumaran: a lot higher for

466 00:41:30.360 ⇒ 00:41:35.249 Uttam Kumaran: right? Like, we’re maybe you’re spending a couple of hours like doing a simple thing. It’s like a kind of

467 00:41:35.740 ⇒ 00:41:38.029 Uttam Kumaran: just a waste. It’s like inefficient.

468 00:41:40.770 ⇒ 00:41:43.330 brian: Yeah, the fact that you’re selling kind of like

469 00:41:43.590 ⇒ 00:41:48.779 brian: products almost now instead of ours. I think, is a way.

470 00:41:50.770 ⇒ 00:41:57.659 brian: a week out of hiring you either hire like expensive smart people, or or cheap.

471 00:41:57.660 ⇒ 00:42:03.469 Uttam Kumaran: Yeah, so what we did is like, do the hourly thing. It was just tough, like you get caught in like a spin cycle of like

472 00:42:03.840 ⇒ 00:42:10.639 Uttam Kumaran: expectations and like, what’s your hours? And then in data, it’s not very linear, like hours to work.

473 00:42:12.410 ⇒ 00:42:15.130 Uttam Kumaran: you know, like 3 h of Dvt work.

474 00:42:15.440 ⇒ 00:42:22.740 Uttam Kumaran: It’s tough to say what you get out of it. So instead, what we do is we, we have like a 15 KA month package, and we have a 25 K month package.

475 00:42:23.570 ⇒ 00:42:29.890 Uttam Kumaran: We I, basically we. I’ve done some math internally that says, cool 15 k, you probably get like 10 h of

476 00:42:30.020 ⇒ 00:42:34.090 Uttam Kumaran: an ae work per week, and you get like 15 h of an analyst

477 00:42:34.530 ⇒ 00:42:36.340 Uttam Kumaran: at the 25 k. It’s like

478 00:42:36.460 ⇒ 00:42:49.769 Uttam Kumaran: a little bit more that gives that gives a rough estimate of like, okay? And then what I do is I size a project. So this is why these days we do. We do audits for every project where I basically go in and say, like, Okay, cool, here’s their timeline.

479 00:42:50.240 ⇒ 00:42:52.449 Uttam Kumaran: They’re not. These guys aren’t in a rush.

480 00:42:52.760 ⇒ 00:42:58.579 Uttam Kumaran: This is sort of the scope. It looks like a bunch of Dbt models, some infra setup and then a bunch of analysis work.

481 00:42:58.860 ⇒ 00:43:05.459 Uttam Kumaran: Roughly, we’re gonna I think this is a 15 K project so I could shove into that bucket and be like, it’s 15 KA month.

482 00:43:06.150 ⇒ 00:43:08.589 Uttam Kumaran: There’s gonna be some that are maybe a little that like.

483 00:43:08.900 ⇒ 00:43:16.559 Uttam Kumaran: And on on 15 KA month. We always make money like there is a margin meaning where our goal is for 50% margin.

484 00:43:16.770 ⇒ 00:43:23.609 Uttam Kumaran: So which is like. Probably I would say, accenture’s

485 00:43:24.090 ⇒ 00:43:27.570 Uttam Kumaran: like I would say, that’s low like.

486 00:43:27.830 ⇒ 00:43:36.370 Uttam Kumaran: and you may for me, when I look at that, I’m like, Oh, damn 50% margin. That’s great. But the expenses of the company are are really high, like payroll payroll is 50%.

487 00:43:37.190 ⇒ 00:43:43.760 Uttam Kumaran: So that’s gross margin. Meaning that’s before software. That’s before tax that’s before, like

488 00:43:44.160 ⇒ 00:43:50.909 Uttam Kumaran: fucking zoom and like insurance and shit like that. So for me, it’s like, I want to optimize that.

489 00:43:51.450 ⇒ 00:44:06.329 Uttam Kumaran: And it’s also steady. So for clients, they’re like cool. It’s 15 K. Because if we do hourly sometimes it’s like 12 k, sometimes 18 k, and so I can just put people on the 15 K subscription, and then when they get maybe 2 beyond that, I could be like, cool. You guys should move to 25 K,

490 00:44:07.160 ⇒ 00:44:12.080 Uttam Kumaran: and then well, I think we’ll start trying to think of packages like that more than we do

491 00:44:12.200 ⇒ 00:44:15.370 Uttam Kumaran: some people. Still, they’re like, Hey, if if maybe we’re

492 00:44:15.890 ⇒ 00:44:22.530 Uttam Kumaran: we don’t need everything we just want like 20 h your time, and then I’ll just charge them at an hourly rate. That’s higher than

493 00:44:22.840 ⇒ 00:44:27.689 Uttam Kumaran: like for 15 K. Because of the they, they actually get a lower net rate.

494 00:44:28.080 ⇒ 00:44:31.070 Uttam Kumaran: But what we benefit from is the stability.

495 00:44:31.210 ⇒ 00:44:36.220 Uttam Kumaran: I know they’re gonna be paying 15 KI know it’s coming versus like

496 00:44:36.550 ⇒ 00:44:38.920 Uttam Kumaran: having to do the dance of ours.

497 00:44:39.640 ⇒ 00:44:41.090 Uttam Kumaran: That’s sort of where we’re at.

498 00:44:43.550 ⇒ 00:44:50.839 brian: Yeah, thinking back to your like previous problem. It’s like.

499 00:44:51.800 ⇒ 00:44:54.409 brian: yeah. So you’re selling packages. That’s good.

500 00:44:55.570 ⇒ 00:44:57.589 brian: I think you just need to.

501 00:44:58.350 ⇒ 00:45:02.610 brian: You need to take a look at like where you’re where not not to tell you how to run.

502 00:45:02.610 ⇒ 00:45:03.819 Uttam Kumaran: No, no, please.

503 00:45:04.010 ⇒ 00:45:10.649 brian: But but like, just look at like where your hours are going. By your like lower skilled employees. Right.

504 00:45:10.650 ⇒ 00:45:11.170 Uttam Kumaran: Yes.

505 00:45:11.170 ⇒ 00:45:24.670 brian: And you’re you’re trying to. Basically, I think you’re, you’re doing this. You’re doing this. But like you just augment it with AI products so that they they can take on more, more products. But the problem is like, if you tell them like, if you ask them, what can I?

506 00:45:25.020 ⇒ 00:45:30.480 brian: What kind of AI agents can I do to use to make your job easier? They they don’t. They can’t think of that.

507 00:45:30.480 ⇒ 00:45:31.839 Uttam Kumaran: I don’t ask, but also.

508 00:45:31.840 ⇒ 00:45:35.020 brian: You have to. You have to figure that out and just have someone build it for you right.

509 00:45:35.020 ⇒ 00:45:39.590 Uttam Kumaran: No, we. So we have. We have 2. So we have 2 people on the team that are pure. AI.

510 00:45:39.710 ⇒ 00:45:43.729 Uttam Kumaran: We are actually doing this, but for me do it. The company is an engineering problem

511 00:45:44.240 ⇒ 00:45:47.910 Uttam Kumaran: like, I don’t look at the company like a

512 00:45:48.440 ⇒ 00:45:52.399 Uttam Kumaran: business. It’s a factory. It’s a widget factory.

513 00:45:53.020 ⇒ 00:45:57.660 Uttam Kumaran: So for me, I’m like, where can we plug AI in right like

514 00:45:57.830 ⇒ 00:46:02.649 Uttam Kumaran: I. And the nice thing is, I’m a great use case because I know how to do the data work.

515 00:46:03.010 ⇒ 00:46:10.409 Uttam Kumaran: So then, I look at like, how would I automate me? Not on the analysis side, because dude part of a lot of our work is relationship building.

516 00:46:11.070 ⇒ 00:46:14.270 Uttam Kumaran: The analysis side is where we actually interact with clients

517 00:46:14.510 ⇒ 00:46:21.520 Uttam Kumaran: automating that you may think is like low hanging fruit. But in fact, it’s probably most of why we even win business.

518 00:46:21.870 ⇒ 00:46:38.490 Uttam Kumaran: and I like to think that we’re better than all the competition, and I do think that we’re pretty good. But most likely there’s there’s a there’s actually a big emotional element to it that is interacting with clients explaining data, explaining decisions that I do think that people on the engineering side are really

519 00:46:38.910 ⇒ 00:46:43.060 Uttam Kumaran: are really like not considering.

520 00:46:43.180 ⇒ 00:46:46.680 Uttam Kumaran: I think, the people that are heavy on the data side. They think

521 00:46:47.330 ⇒ 00:46:53.550 Uttam Kumaran: they? They think that the whole, the whole stack, is an engineering problem, when, in fact, there’s a huge human element

522 00:46:53.750 ⇒ 00:46:59.450 Uttam Kumaran: to to doing, reporting right, because ultimately our job is to help people make decisions.

523 00:46:59.640 ⇒ 00:47:11.039 Uttam Kumaran: And the key word, there is people we’re not like handing this to another agent right yet. There’s no like CEO agent. I’m like interacting with. It’s like a it is a person. So

524 00:47:11.470 ⇒ 00:47:18.980 Uttam Kumaran: I’m but the the thing is dude for the most of our clients. They don’t give a fuck about dbt. They don’t give a fuck about 5 tran, and they

525 00:47:19.490 ⇒ 00:47:22.399 Uttam Kumaran: they also may not care about the warehouse.

526 00:47:23.960 ⇒ 00:47:24.470 brian: Yeah.

527 00:47:24.470 ⇒ 00:47:25.280 Uttam Kumaran: So

528 00:47:26.660 ⇒ 00:47:32.509 Uttam Kumaran: I’m telling you like they don’t even they don’t know. It’s not that they don’t care. They’re like you handle handle it. I don’t care.

529 00:47:32.930 ⇒ 00:47:33.930 Uttam Kumaran: I get you can.

530 00:47:33.930 ⇒ 00:47:35.350 brian: They don’t care about the details. Yeah.

531 00:47:35.350 ⇒ 00:47:40.110 Uttam Kumaran: It’s not. It’s not important. Important is that the data is on time, and it’s full, and

532 00:47:40.640 ⇒ 00:47:45.300 Uttam Kumaran: it has the stuff they need. So. But I don’t know. I think that could be an interesting project.

533 00:47:46.980 ⇒ 00:47:57.120 Uttam Kumaran: You know that I want to start to do that sort of between data platform and AI, the other stuff I was, you know, I thought a lot about sort of like platform stuff. And I messaged about like.

534 00:47:57.330 ⇒ 00:48:02.590 Uttam Kumaran: I’m thinking more about like iceberg and Doc TV. And

535 00:48:03.280 ⇒ 00:48:10.420 Uttam Kumaran: I don’t know. I just don’t think we have. Like the clients are. We have are pretty happy with just using a normal data warehouse.

536 00:48:11.710 ⇒ 00:48:12.640 Uttam Kumaran: I think.

537 00:48:12.640 ⇒ 00:48:14.650 brian: What’s what’s that? Snowflake? Normal data.

538 00:48:14.650 ⇒ 00:48:24.395 Uttam Kumaran: Or people usually have bigquery, and we don’t have like, if we don’t have the ability to change that, then. Yeah, it’s usually one of those, I think some, I think.

539 00:48:25.610 ⇒ 00:48:28.760 Uttam Kumaran: this is also where over time, I want us to get more technical.

540 00:48:29.140 ⇒ 00:48:32.449 Uttam Kumaran: like, I want us to actually go after projects that are like

541 00:48:32.880 ⇒ 00:48:42.739 Uttam Kumaran: new technologies, bigger migrations, like the stuff we’re doing now, I think, is good. But for us to be able to go after like larger contracts.

542 00:48:42.900 ⇒ 00:48:47.329 Uttam Kumaran: That’s where I’m thinking, like, what are the up and coming technologies that could be good to learn

543 00:48:47.460 ⇒ 00:48:53.179 Uttam Kumaran: and like sort of like, start to try to do some stuff internally on or like build on

544 00:48:53.650 ⇒ 00:48:56.790 Uttam Kumaran: iceberg, I think, is one of them where I’m like really curious.

545 00:48:57.913 ⇒ 00:49:04.339 Uttam Kumaran: How it works, and like its effectiveness of moving data across like various platforms.

546 00:49:04.740 ⇒ 00:49:09.500 Uttam Kumaran: I think also, I’ve been looking at Dlt. It’s sort of like Etl as code.

547 00:49:09.980 ⇒ 00:49:13.169 Uttam Kumaran: That’s somewhere. I think that can help us a lot, because

548 00:49:14.043 ⇒ 00:49:17.179 Uttam Kumaran: for every client we have to go procure 5. Tran.

549 00:49:17.370 ⇒ 00:49:22.379 Uttam Kumaran: Set it up, then wait for data to load. It’s probably like a 3 week process.

550 00:49:23.260 ⇒ 00:49:27.809 Uttam Kumaran: Instead, if we have a lot of code already written for data movement.

551 00:49:28.350 ⇒ 00:49:32.359 Uttam Kumaran: and then it’s really just consuming the Api keys, or whatever the auth methods are.

552 00:49:32.490 ⇒ 00:49:34.149 Uttam Kumaran: In order to make that happen

553 00:49:34.740 ⇒ 00:49:40.280 Uttam Kumaran: and dlt. They’re basically trying to do what airbite was trying to do except before air by sort of like, raised a bunch of money. And

554 00:49:40.900 ⇒ 00:49:44.140 Uttam Kumaran: everybody I talked to about using that product said, it sucks so.

555 00:49:44.980 ⇒ 00:49:50.520 Uttam Kumaran: And yeah, that’s what I’m thinking on the data platform side. More, it’s like.

556 00:49:50.790 ⇒ 00:49:52.760 Uttam Kumaran: what are the tougher technologies that

557 00:49:53.510 ⇒ 00:49:58.320 Uttam Kumaran: people are investing in that are really tough de problems that maybe we can try to hook onto.

558 00:50:00.940 ⇒ 00:50:04.980 brian: Yeah, there’s definitely like a lot of, there’s a lot of like

559 00:50:05.230 ⇒ 00:50:09.569 brian: data products and stuff like that. I don’t know, like, I guess the approach I would take is kind of like.

560 00:50:09.720 ⇒ 00:50:14.460 brian: know what all this stuff is, and what it can do to help you, and then

561 00:50:15.080 ⇒ 00:50:18.300 brian: just don’t try to like, make it fit into like

562 00:50:18.930 ⇒ 00:50:22.369 brian: a problem that you don’t have. Kind of. If your problem is just like.

563 00:50:22.910 ⇒ 00:50:26.179 brian: get dbt up as fast as you can. Then you know.

564 00:50:26.180 ⇒ 00:50:26.790 Uttam Kumaran: That’s the price.

565 00:50:26.790 ⇒ 00:50:30.430 brian: Expert is is nice, but like it’s not gonna really help that much.

566 00:50:30.430 ⇒ 00:50:34.669 Uttam Kumaran: I guess I’m trying to think about if any of those new technologies are actually like

567 00:50:35.050 ⇒ 00:50:41.480 Uttam Kumaran: good, not about using them for the existing clients, but getting good at them, and then learning like how to

568 00:50:41.940 ⇒ 00:50:44.920 Uttam Kumaran: how to go after like clients that need that sort of work.

569 00:50:45.670 ⇒ 00:50:49.869 Uttam Kumaran: For example, I doubt there’s a lot of consultancies that that know how how iceberg works.

570 00:50:50.640 ⇒ 00:50:52.170 Uttam Kumaran: But I want to know like

571 00:50:52.790 ⇒ 00:50:57.009 Uttam Kumaran: is that because there’s no clients that need it? Or is that because

572 00:50:57.340 ⇒ 00:50:59.979 Uttam Kumaran: it’s just early, right like that’s my question.

573 00:51:00.990 ⇒ 00:51:08.739 brian: Yeah, I think it’s just like data products before it’s time. It’s like, Dbt, 10 years ago, right? People would be like, why, why do I need this? But now everyone’s on it. I think

574 00:51:08.850 ⇒ 00:51:10.920 brian: just writing whatever is hot is

575 00:51:11.540 ⇒ 00:51:20.019 brian: I. I feel like in data is like the way the way to go, because then you don’t have to do the work of trying to convince them like you need iceberg with a

576 00:51:21.056 ⇒ 00:51:25.569 brian: whatever the the semantic layers on top of that and stuff like that right?

577 00:51:26.430 ⇒ 00:51:32.500 brian: If if they’re just like dbt, dbt is what best practice right now, and just give them dbt.

578 00:51:33.440 ⇒ 00:51:36.480 brian: and just like, get really good at spinning up dvt like.

579 00:51:36.480 ⇒ 00:51:41.379 Uttam Kumaran: Yeah, and dude. We don’t even use cloud much at all.

580 00:51:43.260 ⇒ 00:51:49.269 brian: So what’s your like? Typical, like the the ones you just signed for, like 15 K. Or whatever a month like, what is.

581 00:51:50.060 ⇒ 00:51:57.570 brian: what is the work there, I guess. Is it like data movement into like a new warehouse, and then dbt, work on top of that, and then, like

582 00:51:57.930 ⇒ 00:52:01.430 brian: Bi, on top of that, or like where the stack, usually like, come in.

583 00:52:01.430 ⇒ 00:52:08.879 Uttam Kumaran: Yeah, like, we come in at 2 places, one, sometimes client just want us to do like event modeling.

584 00:52:09.110 ⇒ 00:52:21.190 Uttam Kumaran: And then basically, like, we go in and we help them decide their front end events. And then they’re they’re basically like we, it unlocks a couple of different, bigger problems. Typically, it’s just dbt a warehouse, and then Bi.

585 00:52:21.540 ⇒ 00:52:25.880 Uttam Kumaran: sometimes it’s sometimes it’s no bi.

586 00:52:27.930 ⇒ 00:52:33.780 brian: So, okay, so the 1st one Dbc or so they have like events on like Google, analytics or something. And they they.

587 00:52:33.780 ⇒ 00:52:41.300 Uttam Kumaran: Yeah, channel segments. And then they’re like, Hey, we need help, basically like tracking these and having a strategy around tracking them.

588 00:52:41.570 ⇒ 00:52:43.820 Uttam Kumaran: But then, of course, what that does is like,

589 00:52:45.850 ⇒ 00:52:48.676 Uttam Kumaran: of course, that’s like, we need to decide.

590 00:52:50.890 ⇒ 00:52:56.320 Uttam Kumaran: we need to basically decide like who’s consuming that data. And then we end up finding out that like, okay, you guys need a warehouse.

591 00:52:56.560 ⇒ 00:53:12.249 Uttam Kumaran: And then you guys want to link your other data. Sometimes we come in where it’s basically coming in at a point where they’re at some level of data maturity, where the data actually matters to them. And they have budget. And they have urgency. Those are the kind of things that we’re looking for. Most of their clients are like.

592 00:53:13.320 ⇒ 00:53:16.889 Uttam Kumaran: either between a couple 1 million in revenue or like 50 million.

593 00:53:19.820 ⇒ 00:53:20.540 brian: Got it.

594 00:53:20.940 ⇒ 00:53:26.269 brian: Okay? Yeah. Then, if that’s the target client, if that’s the like normal type of work, then just

595 00:53:26.420 ⇒ 00:53:29.670 brian: build more like widget making machines or widgets to.

596 00:53:29.670 ⇒ 00:53:32.680 Uttam Kumaran: But I guess, like maybe make make that process super efficient.

597 00:53:32.860 ⇒ 00:53:41.590 Uttam Kumaran: Setting up setting up. Dbt, is not that hard like?

598 00:53:42.710 ⇒ 00:53:45.670 Uttam Kumaran: Could I get AI to do it, sure, but it just took me an hour.

599 00:53:46.130 ⇒ 00:53:47.499 Uttam Kumaran: Took me 30 min.

600 00:53:48.070 ⇒ 00:53:49.059 Uttam Kumaran: What is it? It’s like.

601 00:53:49.810 ⇒ 00:53:52.180 brian: Then that’s that part is not worth.

602 00:53:52.180 ⇒ 00:53:55.840 Uttam Kumaran: For me. It’s like I wanna slick the wheels on the stuff that are like

603 00:53:56.460 ⇒ 00:53:59.620 Uttam Kumaran: actually a little bit like challenging like Pr reviews.

604 00:54:00.200 ⇒ 00:54:03.869 Uttam Kumaran: And then also I want to get AI to write some of the small changes

605 00:54:04.740 ⇒ 00:54:08.740 Uttam Kumaran: cause. Then the analysts can start doing Dbt work without ever knowing Dbt.

606 00:54:12.920 ⇒ 00:54:15.350 brian: We also need someone to review it, though that knows Dbc.

607 00:54:16.070 ⇒ 00:54:19.879 Uttam Kumaran: But that’s the thing that’s a bottleneck like, and that’s a common point of delay.

608 00:54:26.530 ⇒ 00:54:29.719 Uttam Kumaran: Or how, or or what do you think about it? Writing the Dbt code.

609 00:54:31.410 ⇒ 00:54:35.240 brian: Yeah, I thought, that’s what we’re talking about. Like ticket comes in. It writes the Dbt code. And then.

610 00:54:35.582 ⇒ 00:54:36.950 Uttam Kumaran: Use it? Yeah, yeah.

611 00:54:36.950 ⇒ 00:54:38.149 brian: You got someone to review it.

612 00:54:38.150 ⇒ 00:54:39.039 Uttam Kumaran: That’s fine! That’s fine!

613 00:54:39.040 ⇒ 00:54:39.990 brian: That’s the end goal.

614 00:54:39.990 ⇒ 00:54:43.600 Uttam Kumaran: Taking care of one side of the the thing.

615 00:54:43.600 ⇒ 00:54:47.360 brian: Yeah. Cause, because then you can hire like a Dbc expert to.

616 00:54:47.640 ⇒ 00:54:54.959 brian: you know, go through like 50 Prs a day. It does like reviewing doesn’t take that long as long as it’s like, mostly right.

617 00:54:54.960 ⇒ 00:54:58.320 Uttam Kumaran: Yeah. And also like, for example, if we get good at breaking

618 00:54:58.850 ⇒ 00:55:02.930 Uttam Kumaran: breaking down engagements into like requirements of phases.

619 00:55:03.300 ⇒ 00:55:06.490 Uttam Kumaran: kind of just like ticket, ticket, ticket, ticket, and like.

620 00:55:07.570 ⇒ 00:55:13.110 brian: Yeah. And then, like, then I don’t know, or whoever is. They’re writing the tickets for you, right?

621 00:55:13.260 ⇒ 00:55:15.750 brian: Because for the requirements kind of.

622 00:55:15.750 ⇒ 00:55:20.310 Uttam Kumaran: We’re still doing. I would say, we’re doing a lot of that ourselves, like we almost come.

623 00:55:21.000 ⇒ 00:55:26.799 Uttam Kumaran: We almost come in as like the whole data team. Basically. So you know, we get like, and then we have to like figure out.

624 00:55:27.730 ⇒ 00:55:30.979 brian: Gotcha. Well, then, like a a Pm. To sit in meetings and gather requirements

625 00:55:32.750 ⇒ 00:55:48.990 brian: so super cheap compared to like an engineer Dvt code. So then you just have bad that you have, like one engineer reviewing all of your all of your tickets, all the Prs that the the bonds turning out. So yeah, I think, yeah, that’s what you talked about focusing on. And that’s

626 00:55:49.100 ⇒ 00:55:52.890 brian: where the majority of the time is that don’t spend all this time figuring out like.

627 00:55:53.140 ⇒ 00:56:00.530 brian: how do I automate the the Dbt platform? If it takes you an hour to do it for a client, and you have 5 clients, right? Or something like that. So so, yeah.

628 00:56:01.280 ⇒ 00:56:03.329 Uttam Kumaran: Yeah. And also, that’s a 1 time setup.

629 00:56:04.580 ⇒ 00:56:05.819 brian: Yeah, exactly.

630 00:56:06.000 ⇒ 00:56:08.890 Uttam Kumaran: Just look at where your hours are going, and then just fix that.

631 00:56:08.890 ⇒ 00:56:13.059 Uttam Kumaran: The other thing that’s interesting is automating sort of like a lot of the stuff that you were working on, which is like

632 00:56:13.320 ⇒ 00:56:17.000 Uttam Kumaran: the snowflake snowflake Ops.

633 00:56:20.433 ⇒ 00:56:23.470 brian: That I was working on what the the fema stuff, or what.

634 00:56:23.470 ⇒ 00:56:29.279 Uttam Kumaran: No, no, like the creating tables like grants stuff like that.

635 00:56:30.030 ⇒ 00:56:31.450 Uttam Kumaran: But even that, like.

636 00:56:31.580 ⇒ 00:56:38.320 Uttam Kumaran: yeah, I don’t know. It’s tough, because I handle a lot of that now, and we have scripts for some stuff like creating new roles.

637 00:56:38.680 ⇒ 00:56:40.630 Uttam Kumaran: that sort of stuff where it’s like.

638 00:56:40.860 ⇒ 00:56:43.299 Uttam Kumaran: I guess we should just use like a

639 00:56:44.000 ⇒ 00:56:48.749 Uttam Kumaran: tool like the one that you were working on, where you basically put in stuff, and it spits out all the

640 00:56:49.290 ⇒ 00:56:50.820 Uttam Kumaran: SQL code to run.

641 00:56:53.230 ⇒ 00:56:57.179 brian: Oh, yeah, yeah, that was so long ago. I forgot about that project

642 00:56:57.440 ⇒ 00:56:59.209 brian: that was like the building for a

643 00:57:02.270 ⇒ 00:57:07.240 brian: But why do you even care about grants doesn’t like all of your doesn’t all your people have just access to everything.

644 00:57:08.420 ⇒ 00:57:11.629 Uttam Kumaran: Well, I well, we come in. And we implement role based access.

645 00:57:13.620 ⇒ 00:57:14.233 brian: Oh, for

646 00:57:14.540 ⇒ 00:57:19.040 Uttam Kumaran: Some people have some people don’t have any. Some people are like we have a fucking warehouse with shit in it.

647 00:57:21.210 ⇒ 00:57:33.789 Uttam Kumaran: Dude, I’m telling you. And it’s brutal. And these guys are making a, these guys are making a shit load of money. So it’s like some we come in. And they’re like, yeah, we have a. We have a couple of views here, tables here all the Ross, that’s all in one area. So my cool, you need governance.

648 00:57:34.040 ⇒ 00:57:40.210 Uttam Kumaran: We need. Dbt, we need some sort of observability alerting like we have to do the whole thing.

649 00:57:45.070 ⇒ 00:57:47.650 brian: Yeah. You know, I was looking at like snowflake

650 00:57:47.760 ⇒ 00:57:56.160 brian: jobs, like people hiring snowflake engineers to see what jobs, what people need. All of them need like our back stuff like this. Our back thing is still not a sold.

651 00:57:56.643 ⇒ 00:57:59.860 Uttam Kumaran: Yo! What great scripts for this! Now.

652 00:58:01.490 ⇒ 00:58:04.790 brian: But yeah, that’s that’s what you’re talking about. Right then, then.

653 00:58:04.790 ⇒ 00:58:09.840 Uttam Kumaran: I mean, we come in and we, we have, like, we create all the roles we create the

654 00:58:10.230 ⇒ 00:58:13.529 Uttam Kumaran: schemas, and then we do all the grants.

655 00:58:14.327 ⇒ 00:58:16.830 Uttam Kumaran: We create roles for service account roles.

656 00:58:17.840 ⇒ 00:58:23.479 brian: Right then it’s a solved problem. I thought you said you’re you’re like trying to figure out how to automate that stuff. It sounds like you already have.

657 00:58:23.480 ⇒ 00:58:27.609 Uttam Kumaran: No, but I guess it’s like it’s like, let’s say, there’s like a slight changes we need to make

658 00:58:29.680 ⇒ 00:58:38.746 Uttam Kumaran: or like. We need to create objects that are a little bit harder like, for example, not. I’m probably only one of the company that knows how to do like pipes and

659 00:58:39.340 ⇒ 00:58:47.039 Uttam Kumaran: storage accounts and snowpipe like that shit is. It’s hard to read the documentation to figure that shit out.

660 00:58:47.740 ⇒ 00:58:53.749 Uttam Kumaran: It would be nice for, like any de to basically have like a co-pilot that helps them. Basically do that.

661 00:58:56.300 ⇒ 00:58:56.900 brian: Yeah.

662 00:58:59.670 ⇒ 00:59:05.319 brian: yeah, this stuff is more niche, though, so hopefully, you don’t have to do it too often. It’s like, how often do you like to snow pipe right.

663 00:59:05.530 ⇒ 00:59:16.309 Uttam Kumaran: It’s more about like, Yeah, you’re you’re right. I mean, it’s just it’s just that over time. I wanna get to more of the complicated stuff like, I don’t wanna just be doing this is the thing

664 00:59:16.710 ⇒ 00:59:30.440 Uttam Kumaran: me saying it’s it’s like the Dbc stuff is pretty standard. That means that there’s gonna be more competition, like, I’m not the only one doing this sort of stuff and probably sees the opportunity I want to start getting into the more Comp, I want to get to like

665 00:59:30.570 ⇒ 00:59:37.909 Uttam Kumaran: the real tough fucking problems, large scale data, multi cloud workloads like

666 00:59:39.210 ⇒ 00:59:45.949 Uttam Kumaran: stuff that’s like, actually a challenge. Because that’s gonna take months like some of those could be like 6 months, 9 months

667 00:59:46.340 ⇒ 00:59:51.680 Uttam Kumaran: contracts they need like 10 people. That’s the sort of stuff that’s really gonna move the needle

668 00:59:57.670 ⇒ 01:00:01.309 Uttam Kumaran: because there’s not a lot of consulting firms that can that can actually do that sort of stuff.

669 01:00:02.100 ⇒ 01:00:09.560 brian: There’s almost like a premature optimization, though it’s kind of like you’re trying to solve for those problems when you haven’t got those contracts. I see your point being like.

670 01:00:09.560 ⇒ 01:00:09.990 Uttam Kumaran: More like.

671 01:00:10.348 ⇒ 01:00:17.880 brian: Someone can come in and do cheaper dbt, modeling, if like, and can like kind of take it right. But like

672 01:00:18.040 ⇒ 01:00:22.069 brian: it’s harder for them to take a 10 person contract when you’re working on a really hard problem.

673 01:00:23.060 ⇒ 01:00:23.590 Uttam Kumaran: It’s

674 01:00:23.590 ⇒ 01:00:29.407 Uttam Kumaran: it’s I mean one. I have to. Yeah, I’m always thinking about like where we’re gonna go. I think the other problem is

675 01:00:29.890 ⇒ 01:00:30.670 Uttam Kumaran: like.

676 01:00:32.960 ⇒ 01:00:39.640 Uttam Kumaran: I’m always thinking where we’re gonna go. The other big problem is that recruiting is always a problem dude.

677 01:00:39.850 ⇒ 01:00:44.570 Uttam Kumaran: So people become the the bottleneck in my business like

678 01:00:45.010 ⇒ 01:01:00.489 Uttam Kumaran: we’re not. You’re never gonna be able to get all the great people that you need. And so what happens? You have to make sacrifices on quality, and when you make sacrifices on quality, the client gets hurt. So for me, it’s building sort of frameworks. Part of that we do through process.

679 01:01:00.970 ⇒ 01:01:02.849 Uttam Kumaran: Part of that I want to do with AI.

680 01:01:03.010 ⇒ 01:01:07.490 Uttam Kumaran: And frankly, I I think I think some point soon I’ll have AI writing.

681 01:01:07.900 ⇒ 01:01:11.080 Uttam Kumaran: I’ll have AI writing more code than people at the company.

682 01:01:11.990 ⇒ 01:01:13.919 Uttam Kumaran: I don’t think we’re like very far from that.

683 01:01:15.150 ⇒ 01:01:15.740 brian: Hmm.

684 01:01:16.330 ⇒ 01:01:22.609 Uttam Kumaran: I just think it’s like I want to do it first.st I don’t think we see it in the market, not because it’s not possible.

685 01:01:23.190 ⇒ 01:01:24.479 Uttam Kumaran: It’s probably hard.

686 01:01:24.990 ⇒ 01:01:27.290 Uttam Kumaran: It’s probably hard, but also

687 01:01:28.040 ⇒ 01:01:34.899 Uttam Kumaran: not many people are uniquely advantaged, like not many data consultancies also do AI work.

688 01:01:36.440 ⇒ 01:01:39.900 Uttam Kumaran: So there’s no way that they would be able to solve their own problem.

689 01:01:40.090 ⇒ 01:01:45.439 Uttam Kumaran: and none of the AI consultancies have any clue. The fuck up like Dvt is

690 01:01:45.800 ⇒ 01:01:50.369 Uttam Kumaran: like my AI. If you guys dude, they don’t even know anything about the date. They have no idea

691 01:01:51.150 ⇒ 01:01:53.759 Uttam Kumaran: about how what what we do on the data side

692 01:01:54.100 ⇒ 01:01:58.290 Uttam Kumaran: like, and they’ve no clue. They’ve like, never interacted with data stuff. All they do is like

693 01:01:59.320 ⇒ 01:02:05.009 Uttam Kumaran: work without labs, same thing on same thing. On the data side. Nobody there is really like

694 01:02:05.280 ⇒ 01:02:08.160 Uttam Kumaran: they use cursor, or they use copilot. But that’s it.

695 01:02:09.300 ⇒ 01:02:19.780 Uttam Kumaran: That’s why I think there’s opportunity to build, actually build like some great AI powered software, the automate.

696 01:02:20.360 ⇒ 01:02:26.500 Uttam Kumaran: snowflake data engineering and Dbt data modeling work that’s not being built in the market.

697 01:02:29.080 ⇒ 01:02:32.910 Uttam Kumaran: Like, that’s kind of like where I don’t know. I just think we have a like.

698 01:02:33.100 ⇒ 01:02:39.970 Uttam Kumaran: Unfortunately, I don’t have like a crack R&D budget that I can like dedicate and like I can’t spend all my time there.

699 01:02:40.070 ⇒ 01:02:42.060 Uttam Kumaran: So we’re slowly like

700 01:02:42.390 ⇒ 01:02:47.349 Uttam Kumaran: building up to it. The engineering will be the hardest thing we try to automate out of everything.

701 01:02:49.580 ⇒ 01:02:50.210 brian: Yeah.

702 01:02:50.490 ⇒ 01:03:00.700 brian: I mean, these are the same problems everyone is trying to solve right? Not everyone. But like like the Zuckerberg, they’re all everyone is trying to like. I saw Zuckerberg says something like, Oh, they they’re gonna replace like mid.

703 01:03:00.700 ⇒ 01:03:13.370 Uttam Kumaran: Their stuff is so hard like, I don’t think. Tell me that we work like a majority of the Prs majority of the Dvc. Prs, what sort of like? What are they? What’s the length and what are the what’s the complication?

704 01:03:14.570 ⇒ 01:03:16.099 brian: I don’t look at any of that stuff

705 01:03:16.210 ⇒ 01:03:18.159 brian: that’s I’m on the platform, but.

706 01:03:18.160 ⇒ 01:03:30.599 Uttam Kumaran: Okay, okay. But you may poke, or maybe poke and look at or even look at, the platform changes you’re making. I don’t know. I just feel like most of like there’s an 80 20. I don’t think most. I think there’s, I think, a lot of the Prs are pretty simple.

707 01:03:33.870 ⇒ 01:03:36.460 brian: Yeah, probably. But that’s cause like

708 01:03:37.340 ⇒ 01:03:42.580 brian: that’s because everyone that we work still left is coasting. It’s just like no one’s doing any, no one. No one’s doing anything.

709 01:03:42.811 ⇒ 01:03:52.079 Uttam Kumaran: Know. But dude, I just think that like, yeah, I also, I think that once you get but once you get to pretty good data maturity. Most of the problem are analysis problems. It’s not like a lot of modeling.

710 01:03:52.310 ⇒ 01:03:53.730 Uttam Kumaran: It’s new data

711 01:03:54.390 ⇒ 01:04:02.209 Uttam Kumaran: and new analysis. The modeling that stuff is like everybody who does modeling for a living is pretty chill like feel like it’s not that stress.

712 01:04:03.230 ⇒ 01:04:10.560 brian: Yeah, you build it. And it’s there. And then, like, you make tweaks to it. Or you fix some like upstream issues. But yeah, you build it. And it’s there.

713 01:04:16.210 ⇒ 01:04:17.970 Uttam Kumaran: Alright. Well, that’s what we’re working on.

714 01:04:19.240 ⇒ 01:04:23.420 brian: Yeah, that’s that’s that is, that is super cool. You’re definitely doing more more exciting stuff than

715 01:04:24.110 ⇒ 01:04:25.169 brian: what’s going on here.

716 01:04:25.170 ⇒ 01:04:31.379 Uttam Kumaran: Yeah. Well, I’ll tell you. Let me know, like I know. Last time we talked it was you were looking more stuff. That’s like

717 01:04:31.660 ⇒ 01:04:34.599 Uttam Kumaran: Async sort of DI may have

718 01:04:34.760 ⇒ 01:04:37.039 Uttam Kumaran: as we start taking on bigger clients.

719 01:04:37.170 ⇒ 01:04:42.129 Uttam Kumaran: I think we’ll definitely have. We definitely have some of that now, is it.

720 01:04:42.440 ⇒ 01:04:50.270 Uttam Kumaran: Is it like 10 or 20 h of work a week? I don’t know. I don’t think it’s at that level yet. It’s definitely like probably 5 to 10 h of work right now.

721 01:04:50.470 ⇒ 01:04:53.720 Uttam Kumaran: That’s like de sort of stuff that you can do, Async, whenever

722 01:04:56.330 ⇒ 01:05:01.400 Uttam Kumaran: because for me, I’m filling in on all of that right now, I’m cause I yeah.

723 01:05:01.400 ⇒ 01:05:03.470 Uttam Kumaran: I’m not doing much a work anymore.

724 01:05:04.600 ⇒ 01:05:06.940 brian: Is it kind of like snowflake, like the.

725 01:05:06.940 ⇒ 01:05:11.209 Uttam Kumaran: Yeah, it’s like, it’s like, it’s like new pipelines. It’s like, it’s like de stuff.

726 01:05:12.631 ⇒ 01:05:15.199 Uttam Kumaran: and then I I need some help, like

727 01:05:15.330 ⇒ 01:05:17.410 Uttam Kumaran: on the platform side. I need help like

728 01:05:17.700 ⇒ 01:05:20.930 Uttam Kumaran: setting up Linkedin, setting up Cicd.

729 01:05:21.150 ⇒ 01:05:25.330 Uttam Kumaran: setting up like staging environments like boilerplate code for that.

730 01:05:26.270 ⇒ 01:05:29.089 Uttam Kumaran: So definitely like, maybe I don’t know what

731 01:05:29.290 ⇒ 01:05:31.829 Uttam Kumaran: if you have time for that? I actually do need like

732 01:05:32.400 ⇒ 01:05:35.870 Uttam Kumaran: I was gonna do that. I’m the only one here that could that could work on that really.

733 01:05:40.250 ⇒ 01:05:42.919 brian: Okay, so what are some more? More? So linting.

734 01:05:43.260 ⇒ 01:05:47.010 Uttam Kumaran: Basically like every repo we wanna have, like Dvt linking

735 01:05:49.080 ⇒ 01:05:55.939 Uttam Kumaran: in our when we, I want you to kind of audit our like snowflake setup code. Take a look. If there’s anything else we need to do there.

736 01:05:56.270 ⇒ 01:06:04.389 Uttam Kumaran: we’re definitely not doing anything on like where I don’t think we’re creating warehouses or anything. That’s probably something we can do. Second thing is on like the actual Cicd for dbt.

737 01:06:04.850 ⇒ 01:06:10.739 Uttam Kumaran: which is like having it having it do the Dvt compile Dvt run.

738 01:06:10.970 ⇒ 01:06:14.130 Uttam Kumaran: executing that in a staging environment, once

739 01:06:14.400 ⇒ 01:06:20.510 Uttam Kumaran: the Pr gets pushed, it gets pushed to like a production environment creating

740 01:06:20.930 ⇒ 01:06:26.499 Uttam Kumaran: as part of like when we onboard a new client, we create basically playground schemas for every

741 01:06:27.240 ⇒ 01:06:29.850 Uttam Kumaran: every like developer, that sort of stuff.

742 01:06:32.020 ⇒ 01:06:38.550 Uttam Kumaran: Because we’re gonna we’re gonna start to multi, we’re we’re basically like, think about we’re basically the data team.

743 01:06:39.060 ⇒ 01:06:47.779 Uttam Kumaran: We’re the data team for like 5 different companies. And then we have a data platform team that supports that of which it’s me.

744 01:06:48.568 ⇒ 01:06:55.210 Uttam Kumaran: Like, I’m the only person thinking about like doing this over and over again. So like, it’s like.

745 01:06:55.960 ⇒ 01:07:02.309 Uttam Kumaran: it’s like data platform on, on like steroids. Because that’s just it’s like way, too. It’s like a lot. But at the same time

746 01:07:02.940 ⇒ 01:07:10.750 Uttam Kumaran: it’s mainly guardrails. It’s mainly like just making sure that people when people come to work, I want them to be able to just develop Dbt code for the most part.

747 01:07:12.860 ⇒ 01:07:13.430 brian: Got it.

748 01:07:13.430 ⇒ 01:07:17.479 Uttam Kumaran: And I want the analyst to like to be able to run queries in a safe way. And things like that.

749 01:07:22.740 ⇒ 01:07:23.250 Uttam Kumaran: Yeah.

750 01:07:26.050 ⇒ 01:07:28.250 brian: I guess where my head’s at is.

751 01:07:28.850 ⇒ 01:07:30.789 brian: I think the snowflake work seems

752 01:07:31.757 ⇒ 01:07:38.440 brian: more up my alley, especially if I’m doing like this in the pro stuff right now. So I’m happy to do that kind of stuff.

753 01:07:41.360 ⇒ 01:07:42.710 brian: Also, if.

754 01:07:43.240 ⇒ 01:07:45.539 Uttam Kumaran: Do you have any interest in the AI stuff.

755 01:07:46.570 ⇒ 01:07:48.065 brian: No, I I do

756 01:07:49.760 ⇒ 01:07:51.730 brian: So for the other platform stuff.

757 01:07:52.390 ⇒ 01:07:54.370 brian: I don’t know how you do like

758 01:07:54.710 ⇒ 01:08:01.449 brian: equity or stuff like that. But if you are looking for someone to kind of like partner on, and then you just want someone to

759 01:08:02.150 ⇒ 01:08:09.340 brian: figure out all your platform stuff, how to make your like, build the platform so that you can spin up and have your have your guys.

760 01:08:10.040 ⇒ 01:08:11.979 brian: you know, be be on a

761 01:08:12.170 ⇒ 01:08:15.090 brian: beyond the like a platform and

762 01:08:16.979 ⇒ 01:08:24.689 brian: Then, like some equity stuff or some partner stuff that I could be more interested in in kind of that that kind of route as well, but I don’t. Yeah.

763 01:08:24.920 ⇒ 01:08:29.320 brian: So that’s kind of where’s my my head’s at for that stuff. But that would that would be more like,

764 01:08:29.850 ⇒ 01:08:32.469 brian: yeah, I’d be like more like fully invested into like.

765 01:08:32.720 ⇒ 01:08:33.310 Uttam Kumaran: Yeah.

766 01:08:33.310 ⇒ 01:08:35.419 brian: You know, 40 h, week type thing, right? And

767 01:08:36.660 ⇒ 01:08:39.490 brian: equity over like salary. But like, that’s yeah.

768 01:08:39.960 ⇒ 01:08:44.710 Uttam Kumaran: Yeah, I, just, we’re just not like all the platform stuff I’m doing is basically like

769 01:08:45.310 ⇒ 01:08:50.140 Uttam Kumaran: as we need it right now, because our number one goal is the building of our client base.

770 01:08:50.390 ⇒ 01:08:58.740 Uttam Kumaran: So right now, our number one priority is sales and basically people, the platform stuff is like

771 01:08:59.420 ⇒ 01:09:04.030 Uttam Kumaran: nice to have, but not our bottleneck right now. That’s the thing.

772 01:09:04.770 ⇒ 01:09:05.580 Uttam Kumaran: However.

773 01:09:05.580 ⇒ 01:09:06.100 brian: Gotcha.

774 01:09:06.109 ⇒ 01:09:08.439 Uttam Kumaran: We are uniquely advantaged because we’re

775 01:09:08.809 ⇒ 01:09:14.259 Uttam Kumaran: we see so many different setups like I’ve

776 01:09:14.629 ⇒ 01:09:18.839 Uttam Kumaran: I’ve just worked for like 15 companies like pretty deeply.

777 01:09:19.115 ⇒ 01:09:19.389 brian: Okay.

778 01:09:19.399 ⇒ 01:09:20.759 Uttam Kumaran: In like a year and a half.

779 01:09:21.575 ⇒ 01:09:25.859 Uttam Kumaran: So we have an opportunity to. Not only

780 01:09:26.089 ⇒ 01:09:31.699 Uttam Kumaran: for me, I think the opportunity really is not on the sit on the platform stuff. It’s I. I really feel like

781 01:09:31.849 ⇒ 01:09:33.949 Uttam Kumaran: we have some opportunity on the AI side.

782 01:09:35.146 ⇒ 01:09:42.849 Uttam Kumaran: Because, of course, there’s gonna be people with better platforms than us, and there’s gonna be some off the shelf platform tooling. But I do think that the AI stuff

783 01:09:43.029 ⇒ 01:09:49.509 Uttam Kumaran: is where, like there’s some real advantage, for for from us, knowing

784 01:09:49.719 ⇒ 01:09:51.949 Uttam Kumaran: what it takes to be a grade de.

785 01:09:52.109 ⇒ 01:10:05.469 Uttam Kumaran: and from us also, I have 2 full time. AI people that are building a shit ton of automation for our business. So like, if you pair with those guys, you guys will rip this whole thing out in like a week.

786 01:10:06.259 ⇒ 01:10:07.789 Uttam Kumaran: This, those guys are cracked.

787 01:10:07.999 ⇒ 01:10:17.609 Uttam Kumaran: They just have no idea. They have no idea what data like, how the fuck, what we, what the data stuff is. So for me, the way I was gonna do is I basically have to just like.

788 01:10:18.309 ⇒ 01:10:23.069 Uttam Kumaran: show them kind of what we need, and then sort of collaborate with them to build it.

789 01:10:23.809 ⇒ 01:10:28.989 Uttam Kumaran: The number one goal there is basically hire less do more faster, you know.

790 01:10:29.585 ⇒ 01:10:33.559 Uttam Kumaran: But I hear you on like the investment side of things like that.

791 01:10:34.404 ⇒ 01:10:39.129 Uttam Kumaran: I would have to think about a longer term, because it would just require

792 01:10:40.489 ⇒ 01:10:46.339 Uttam Kumaran: it’s just a lot because we’re not raising any money. We don’t have plans like on that. So

793 01:10:47.259 ⇒ 01:10:52.549 Uttam Kumaran: for me, I would for me. I’m more inclined just to pay higher, like I would just pay higher.

794 01:10:55.319 ⇒ 01:11:00.169 Uttam Kumaran: But that’s sort of like my gut instinct on that, because, yeah, that’s where we are now.

795 01:11:01.709 ⇒ 01:11:10.799 Uttam Kumaran: But I hear you, too. I hear I hear you, too, and like I also don’t want to have you come. I for me, for me. What’s more important is like, I want you to come and work on like real hard fucking stuff.

796 01:11:11.739 ⇒ 01:11:18.029 Uttam Kumaran: And yeah, that’s actually for me the most top of mind, because otherwise, like

797 01:11:18.559 ⇒ 01:11:21.719 Uttam Kumaran: you should just be chill. I don’t want you to come do normal work, but also.

798 01:11:22.219 ⇒ 01:11:26.579 Uttam Kumaran: if you come work for us, I want to pay you a lot. So

799 01:11:27.939 ⇒ 01:11:32.509 Uttam Kumaran: That’s sort of where I’m at. The opportunity I have now is sort of on this kind of like.

800 01:11:32.619 ⇒ 01:11:37.129 Uttam Kumaran: probably poking around with this AI problem. And then some of the platform stuff that we need.

801 01:11:39.110 ⇒ 01:11:41.440 brian: Yeah. And if, like, the AI stuff is

802 01:11:44.480 ⇒ 01:11:46.120 brian: is kind of like, where.

803 01:11:47.090 ⇒ 01:11:52.130 brian: yeah, I don’t know. Like with the AI stuff, I guess it’s a lot of like it can be a lot of research without like

804 01:11:52.930 ⇒ 01:11:55.539 brian: something coming into fruition. Kind of you know what I mean.

805 01:11:55.540 ⇒ 01:12:00.040 Uttam Kumaran: Oh, no, I’m like Dude. Well, I would fuck in their next client. I would only like

806 01:12:00.520 ⇒ 01:12:04.789 Uttam Kumaran: I wouldn’t put anyone that will be like, let’s run this whole client through. AI, that’s all we do.

807 01:12:05.680 ⇒ 01:12:06.190 brian: Hmm.

808 01:12:06.190 ⇒ 01:12:12.970 Uttam Kumaran: I’m very serious. Yeah, I’m like this. I’m this is not like this is not a research project like we are going to do this

809 01:12:13.190 ⇒ 01:12:18.850 Uttam Kumaran: this year. It’s just matter of when on how fast like.

810 01:12:19.680 ⇒ 01:12:29.100 Uttam Kumaran: If I had do it, if I if I stop everything I would, I would probably have it built by end of February, but it takes time, and it takes like some like.

811 01:12:30.120 ⇒ 01:12:33.350 Uttam Kumaran: It takes some thought into how to do this.

812 01:12:35.460 ⇒ 01:12:36.920 Uttam Kumaran: And I just think that like

813 01:12:37.510 ⇒ 01:12:42.799 Uttam Kumaran: it takes someone at our level who’s seen like kind of plot. These sort of platforms get built and can understand the nuance.

814 01:12:42.910 ⇒ 01:12:48.049 Uttam Kumaran: So that’s sort of what I’m thinking. But no, I mean, I would legitimately just the next client we take on.

815 01:12:48.170 ⇒ 01:12:50.850 Uttam Kumaran: We would try to just do everything via the bot

816 01:12:51.170 ⇒ 01:12:54.809 Uttam Kumaran: and me, or someone at the like the higher engagement level.

817 01:12:55.610 ⇒ 01:12:56.550 Uttam Kumaran: That’s it.

818 01:12:59.491 ⇒ 01:13:14.999 Uttam Kumaran: Well, that’s the nice thing. I don’t have to ask anybody or anything. There’s no bureaucracy like, I’m very serious about like doing that. And then that’s a product where like, for example, this is my head is that if we build an AI agent that does that, and we want to spin that out.

819 01:13:15.310 ⇒ 01:13:19.840 Uttam Kumaran: Then I think that’s a compelling vision for a company like

820 01:13:20.470 ⇒ 01:13:22.699 Uttam Kumaran: that’s what I would form a company around

821 01:13:23.330 ⇒ 01:13:28.110 Uttam Kumaran: and like to be really honest. The consulting thing is tough, like

822 01:13:28.850 ⇒ 01:13:35.110 Uttam Kumaran: the the business of consulting. It’s just a constant sales and constant, just like people.

823 01:13:35.460 ⇒ 01:13:36.930 Uttam Kumaran: If we want to go build a

824 01:13:37.250 ⇒ 01:13:45.460 Uttam Kumaran: a business like that, I would try to just take one of the things that we build either on the platform side or the AI side, find that it works. And then, just like

825 01:13:47.220 ⇒ 01:13:48.990 Uttam Kumaran: spin it out. Basically.

826 01:13:50.910 ⇒ 01:13:51.850 brian: Gotcha.

827 01:13:52.280 ⇒ 01:13:55.030 Uttam Kumaran: Like that’s that’s more. That’s more compelling to me.

828 01:13:55.930 ⇒ 01:14:00.409 brian: That does sound pretty interesting, like, I guess the next

829 01:14:00.630 ⇒ 01:14:09.010 brian: you’re saying so the next client you sign you’re gonna try to do it all through. AI, and you’re gonna kind of like dog food your own your own work to see if it works right.

830 01:14:09.600 ⇒ 01:14:10.190 Uttam Kumaran: Yeah.

831 01:14:10.590 ⇒ 01:14:19.820 brian: And I guess you’re asking if I would be interested in in like working on that project as like like hourly rate type stuff or something right?

832 01:14:19.820 ⇒ 01:14:23.299 Uttam Kumaran: Yeah, I would say, yeah, exactly hourly rate. And then

833 01:14:23.600 ⇒ 01:14:26.280 Uttam Kumaran: I think the hope like to think bigger.

834 01:14:26.410 ⇒ 01:14:27.860 Uttam Kumaran: The hope is.

835 01:14:28.450 ⇒ 01:14:42.650 Uttam Kumaran: I mean one. I think, like you get some money in the short term, I think. Second, my goal is that one of these things we built we build on the platform side, basically can stand up alone and become like a company.

836 01:14:44.390 ⇒ 01:14:45.320 Uttam Kumaran: Right? Yeah.

837 01:14:45.420 ⇒ 01:14:53.620 Uttam Kumaran: I especially think I mean, we’ve been talking about this for the last 2 years like, I still think there’s a shit lot of opportunity in the Snowflake and dbt, like tooling.

838 01:14:54.240 ⇒ 01:14:56.170 Uttam Kumaran: especially in the snowflake tooling.

839 01:14:57.930 ⇒ 01:15:00.129 Uttam Kumaran: I still think there’s a lot of opportunity.

840 01:15:00.590 ⇒ 01:15:11.340 Uttam Kumaran: So for me, I wanna that’s where, like for me, the platform bets are not, partly because I want to help us do things faster. But I also do think that we’re gonna find something that works

841 01:15:11.930 ⇒ 01:15:16.469 Uttam Kumaran: we built. Cause this is building software for ourselves. Once it works, then we’re like.

842 01:15:16.850 ⇒ 01:15:21.909 Uttam Kumaran: what do we need to to sell this? We need a CEO. We need someone to sell this. And then

843 01:15:23.070 ⇒ 01:15:28.040 Uttam Kumaran: we basically reinforce can just do the engineering. So then we spin it out. Someone goes and runs that.

844 01:15:29.420 ⇒ 01:15:30.599 Uttam Kumaran: That’s what I would do.

845 01:15:30.850 ⇒ 01:15:33.720 Uttam Kumaran: The hardest part is understanding. The product is right.

846 01:15:34.070 ⇒ 01:15:40.619 Uttam Kumaran: the sell, the selling shit like that. Bro, that’s a solve problem telling you the hardest part is knowing that the thing works

847 01:15:41.050 ⇒ 01:15:43.969 Uttam Kumaran: like, I really fundamentally believe that, like.

848 01:15:44.540 ⇒ 01:15:47.020 Uttam Kumaran: I think software sales is like

849 01:15:47.670 ⇒ 01:15:51.980 Uttam Kumaran: software sales. Marketing is figured out. Typically most people are marketing something that doesn’t work.

850 01:15:52.550 ⇒ 01:15:57.549 Uttam Kumaran: And they’re building it on the fly. So we would basically incubate it internally if it ends up working.

851 01:15:59.050 ⇒ 01:16:04.588 brian: Yeah. And I I guess, like for for like comp stuff,

852 01:16:05.480 ⇒ 01:16:07.319 brian: I guess what I’m thinking is like.

853 01:16:08.300 ⇒ 01:16:10.180 brian: like, right now, kind of like what

854 01:16:10.690 ⇒ 01:16:12.580 brian: I’m drawn towards this kind of

855 01:16:12.780 ⇒ 01:16:20.689 brian: having ownership kind of like of a product that I work on, or something like that less, even if I don’t get paid like some like.

856 01:16:20.840 ⇒ 01:16:22.450 brian: even if I don’t get paid like an hourly rate

857 01:16:22.970 ⇒ 01:16:25.729 brian: like we were. We were just kind of paying the cash.

858 01:16:26.230 ⇒ 01:16:29.730 Uttam Kumaran: Hear you. I hear you. Yeah, exactly.

859 01:16:29.730 ⇒ 01:16:30.770 brian: My own. Yeah.

860 01:16:30.770 ⇒ 01:16:34.730 Uttam Kumaran: No, I think about what it would take to hire me, and it’s the same thing I mean, like.

861 01:16:35.280 ⇒ 01:16:40.680 Uttam Kumaran: if I was to pay you half a million. Maybe you’d be like, okay, maybe I don’t do that. But I can’t do that.

862 01:16:41.110 ⇒ 01:16:47.499 Uttam Kumaran: But that’s so. Maybe like, think about that like for me. That’s what I would. But so the problem is, I can’t like

863 01:16:47.850 ⇒ 01:16:56.059 Uttam Kumaran: we can sign everything we can, I can promise you, but I don’t know. I don’t have all the answers. All I have is a guinea pig that we can run a lot of experiments on.

864 01:16:57.140 ⇒ 01:17:00.550 Uttam Kumaran: and then it’s kind of like trust at that point.

865 01:17:01.670 ⇒ 01:17:07.590 Uttam Kumaran: cause I don’t. I like I can’t make. I don’t have any. I can’t. I can’t promise anything else, because I don’t know like.

866 01:17:08.710 ⇒ 01:17:11.269 Uttam Kumaran: I don’t know whether it’s good things are gonna work or not, and

867 01:17:11.590 ⇒ 01:17:14.629 Uttam Kumaran: I’m not willing to promise, because there’s no need to, because I

868 01:17:15.270 ⇒ 01:17:22.239 Uttam Kumaran: I’m gonna go. I’m gonna I’m literally gonna go try. I’ll try these things at some point for me. It’s like, Can I stack the deck right like

869 01:17:22.880 ⇒ 01:17:30.209 Uttam Kumaran: for me. The number one thing is, can I get someone that’s as smart as you on board so that we can actually go the distance

870 01:17:30.340 ⇒ 01:17:31.170 Uttam Kumaran: right?

871 01:17:31.640 ⇒ 01:17:39.559 Uttam Kumaran: Cause I can go try to flotz around with this. Maybe it works. But I there’s so much to do. I don’t have the time, and

872 01:17:40.140 ⇒ 01:17:49.429 Uttam Kumaran: like this, the business has to. The 1st thing is, brain forces to keep going for this to even like get incubated right? So for me, it’s like, Can I? How do I attract

873 01:17:49.910 ⇒ 01:17:55.649 Uttam Kumaran: like people like you and a couple of their friends that want to work on something like bigger, and my pitch would be

874 01:17:55.900 ⇒ 01:18:05.969 Uttam Kumaran: look in the short term. It’s it’s cash, which is not nothing in the long term. If we find an opportunity, spin one of these out as a company, I think there’s opportunity there. But

875 01:18:06.790 ⇒ 01:18:09.019 Uttam Kumaran: and I I think that’s a i think that’s a

876 01:18:09.410 ⇒ 01:18:15.699 Uttam Kumaran: that that’s a better guarantee than most. We’ll make it, and you know, and you know me, so I’m not like trying to jippy.

877 01:18:15.700 ⇒ 01:18:16.230 brian: Yeah.

878 01:18:16.230 ⇒ 01:18:17.080 Uttam Kumaran: Just like.

879 01:18:17.080 ⇒ 01:18:17.750 brian: Yeah.

880 01:18:17.750 ⇒ 01:18:20.720 Uttam Kumaran: It’s like, that’s that’s what I can say. That’s that’s my thought.

881 01:18:20.970 ⇒ 01:18:22.280 Uttam Kumaran: So I don’t know. Think about it.

882 01:18:23.040 ⇒ 01:18:30.192 brian: Yeah, when the next projects or contract comes along, and you’re planning on doing this.

883 01:18:31.260 ⇒ 01:18:35.639 brian: I can’t say like, yes, I can commit like this many hours right now, or whatever, just just like kind of loop me in.

884 01:18:35.640 ⇒ 01:18:57.539 Uttam Kumaran: I’ll just have you looped in, and you just poke around at it again, and I’m happy to pay you for it. I don’t want to bring you in on anything where you’re like. Like, if if you if you’re like, Hey, I’m down to do client work, and you’re like, do you have 10 h of client work. There’s stuff like that, if you’re like, you know, that’s not interesting. And more interesting is this stuff. Then, as we sort of have conversations. I’ll just include you. We’re not. We haven’t talked about any of this stuff internally. It’s been all on my head.

885 01:18:58.030 ⇒ 01:19:04.640 Uttam Kumaran: so I’ll just add, I’ll just have. You can just come sit in on chats with me and the AI team as we sort of poke around the problem.

886 01:19:05.170 ⇒ 01:19:07.199 Uttam Kumaran: and that’s it, like no pressure.

887 01:19:08.360 ⇒ 01:19:13.959 brian: Yeah, that sounds good. That’s I think that will give me a better feel of like what’s needed. And then

888 01:19:14.360 ⇒ 01:19:17.159 brian: what? What kind of work is required. Type thing. Okay.

889 01:19:18.640 ⇒ 01:19:19.510 Uttam Kumaran: Okay, the same.

890 01:19:20.840 ⇒ 01:19:22.410 brian: Alright, man, take it easy.

891 01:19:22.410 ⇒ 01:19:26.170 Uttam Kumaran: Oh, yeah, I know we got. I gotta we gotta see each other soon, hopefully. Next time you come back.

892 01:19:27.000 ⇒ 01:19:28.312 brian: Yeah, I need to.

893 01:19:28.760 ⇒ 01:19:31.370 brian: Since you’re you’re you’re my. Remember. You’re my address.

894 01:19:31.370 ⇒ 01:19:31.860 Uttam Kumaran: Yes.

895 01:19:31.860 ⇒ 01:19:38.310 brian: Which is a super solid. So I need to go get my my driver’s license expired, so I need to go to Texas to get my driver’s license.

896 01:19:38.310 ⇒ 01:19:39.760 Uttam Kumaran: Oh, great. Okay.

897 01:19:39.760 ⇒ 01:19:43.300 brian: So so before I do that, I’m probably gonna have them send some like

898 01:19:43.480 ⇒ 01:19:52.739 brian: bank statements or something to your house, because I need that as proof of residence. So I’ll I’ll do that sometime. But I think that I don’t know what that’s gonna be. But yeah, it’s it’s coming up.

899 01:19:53.420 ⇒ 01:19:53.880 brian: Okay.

900 01:19:54.750 ⇒ 01:19:55.900 Uttam Kumaran: Okay, perfect.

901 01:19:57.270 ⇒ 01:19:59.130 Uttam Kumaran: Alright, bye, good chat.

902 01:19:59.130 ⇒ 01:19:59.499 brian: Talk to you.

903 01:19:59.500 ⇒ 01:19:59.860 Uttam Kumaran: But just.

904 01:19:59.860 ⇒ 01:20:02.720 brian: Yeah. Good stuff. Yeah. Talk to you later. Bye.