Meeting Title: Brainforge Engineering Sync Date: 2024-12-10 Meeting participants: Luke Daque, Uttam Kumaran, Payas Parab, Miguel De Veyra, Casie Aviles


WEBVTT

1 00:03:40.790 00:03:41.690 Uttam Kumaran: Hey, guys.

2 00:03:46.030 00:03:46.680 Casie Aviles: Yeah. Done.

3 00:03:47.360 00:03:49.530 Uttam Kumaran: We could stay video off. I don’t mind

4 00:03:51.620 00:03:57.519 Uttam Kumaran: this is not not. It’s not a client call, hey, Pius.

5 00:03:57.670 00:03:59.110 Payas Parab: Hey, Tom, how are you, ma’am.

6 00:03:59.110 00:03:59.830 Uttam Kumaran: Good.

7 00:04:00.060 00:04:07.691 Uttam Kumaran: I said we could stay video off. We could do also do video on. But since it’s not a client call, I feel like I could give everyone some video off time.

8 00:04:07.930 00:04:10.119 Payas Parab: Sure for me.

9 00:04:10.120 00:04:19.430 Uttam Kumaran: Okay, cool. So let me ping Ryan and make sure he can join

10 00:04:30.650 00:04:33.700 Uttam Kumaran: heads up. We told him. I do have to drop off in like 30 min. So.

11 00:04:33.700 00:04:38.170 Uttam Kumaran: okay, that’s fine. Yeah, I think we should be able to get through everything fairly quickly. So

12 00:04:38.844 00:04:43.380 Uttam Kumaran: yeah, maybe let’s just get started. So I wrote up this engineering sync, Doc.

13 00:04:43.837 00:04:47.440 Uttam Kumaran: basically, I just kind of like, wanted to go through a couple of questions

14 00:04:47.590 00:04:53.169 Uttam Kumaran: that I had, for you know, everyone. But also, I think, just use this meeting to kind of chat

15 00:04:53.250 00:05:20.973 Uttam Kumaran: with everyone about like what they’re working on. And if anyone needs any help with anything. I think this meeting, of course, we’re talking about data, and AI stuff. So I think for pie. So I think it’ll be helpful. You can kind of see a little bit about the stuff we’re working on on the AI side. And then I think on for the data for the AI folks to kind of get a sense of the stuff we’re working on the data side. So that’ll be really good. Yeah, I think maybe just to kick it off. I think.

16 00:05:21.260 00:05:25.780 Uttam Kumaran: I wanted just to talk to everyone about

17 00:05:25.840 00:05:29.629 Uttam Kumaran: like how they’re tracking like their daily to do’s

18 00:05:30.030 00:05:39.950 Uttam Kumaran: and I don’t. You know, I think everybody has their own process like my process is insane right now, like, I probably don’t have a process. I basically wake up and

19 00:05:40.410 00:05:44.250 Uttam Kumaran: check slack every 90 seconds for about 8 h.

20 00:05:44.583 00:05:56.899 Uttam Kumaran: But I ideally that’s not everybody’s process. So kind of like my vision, for this one is like it’s kind of what I mentioned last week, which is first.st We just have to make sure all all work is tracked somewhere.

21 00:05:57.267 00:06:03.432 Uttam Kumaran: And so I think maybe it’ll just kind of like, go around the Horn. We don’t have to share anything like on screen or anything. But

22 00:06:03.910 00:06:11.670 Uttam Kumaran: Maybe Casey Miguel, like, you guys want to go and just talk about like where your work is tracked every day, and how you kind of feel about like

23 00:06:12.030 00:06:14.629 Uttam Kumaran: where we are currently with like tracking everything.

24 00:06:17.240 00:06:19.390 Miguel de Veyra: Yeah, I guess I’ll start.

25 00:06:19.770 00:06:23.790 Miguel de Veyra: As for my tracking honestly.

26 00:06:24.960 00:06:26.849 Miguel de Veyra: I just ask you, Uta, on which.

27 00:06:26.850 00:06:37.239 Uttam Kumaran: And you can be honest. Yeah, be totally honest, because where we are now is not where we’ll be in the future. So this is more just getting a sense of where we are today. So don’t worry if you’re like, I just have a piece of paper. That’s fine.

28 00:06:37.240 00:06:42.720 Miguel de Veyra: Yeah, honestly, what I do is, I just depend entirely on that AI automations thing.

29 00:06:43.110 00:06:56.590 Miguel de Veyra: the board we have like the in notion. Then I just ask you on, you know which is top priority, or I just check the calendar which Demo is the closest, or which Demo is like most likely for us to close. And then that’s what I work on.

30 00:06:57.965 00:07:00.189 Miguel de Veyra: I guess that was my

31 00:07:00.210 00:07:06.870 Miguel de Veyra: general process over the past few months, though it kind of changed since you told me to track our hours on Vitaco.

32 00:07:07.200 00:07:07.780 Uttam Kumaran: Yeah.

33 00:07:08.300 00:07:11.640 Miguel de Veyra: So sometimes, like I switch on and off, you know.

34 00:07:12.099 00:07:15.420 Miguel de Veyra: they’re on crocify, because that’s where we track everything.

35 00:07:16.200 00:07:22.650 Uttam Kumaran: Vitaco work isn’t really tracked like in terms of what tasks are up next, and stuff like that.

36 00:07:24.528 00:07:27.750 Miguel de Veyra: No, no, not really cause. Right now we’re debugging.

37 00:07:30.270 00:07:34.039 Uttam Kumaran: But even like that debugging task there, it’s not. It’s just not anywhere in ocean.

38 00:07:35.030 00:07:37.649 Miguel de Veyra: No, no, no, it’s not. It’s more than clockify.

39 00:07:39.670 00:07:40.250 Uttam Kumaran: Cool.

40 00:07:41.070 00:07:50.640 Miguel de Veyra: Because ideally, I’m not sure if it’s possible to do it. Because, remember, Casey, we had this thing before in Asana, where there’s like boards for each project.

41 00:07:51.710 00:07:52.210 Casie Aviles: Yeah.

42 00:07:52.210 00:08:02.909 Miguel de Veyra: So ideally, we ideally, I want to implement something like that when we have more clients on the AI side. But I don’t think it’s necessary right now, when we only have, you know, basically only

43 00:08:04.510 00:08:05.770 Miguel de Veyra: vitaco.

44 00:08:10.220 00:08:13.129 Miguel de Veyra: But yeah, I think that’s pretty much it on my end.

45 00:08:13.690 00:08:17.200 Uttam Kumaran: Okay, what about you, Casey? Like? How is how like when you wake up?

46 00:08:17.490 00:08:21.920 Uttam Kumaran: What are you checking 1st and like, how are you kind of arranging your day.

47 00:08:22.740 00:08:26.770 Casie Aviles: Yeah. So for me, I typically check slack first.st

48 00:08:26.990 00:08:33.230 Casie Aviles: And when I when I started I didn’t, you know I didn’t check the notion board yet.

49 00:08:33.419 00:08:37.030 Casie Aviles: But you know, lately, like I get that

50 00:08:37.080 00:08:41.970 Casie Aviles: we’re trying to track our tasks better. So I was

51 00:08:42.179 00:08:52.150 Casie Aviles: slowly getting into notion. And yeah, and I also think it’s also better, because there’s like a visual. I’m more used to kanban boards personally. So

52 00:08:52.540 00:08:58.360 Casie Aviles: yeah, so that’s why I think okay, I should check notion much more. And then I also started.

53 00:08:58.972 00:09:02.380 Casie Aviles: You know, inputting some, adding some

54 00:09:02.410 00:09:11.610 Casie Aviles: cards there for the stuff that I’m working on. And yeah, and also we sometimes I get on a call with Miguel, and he also

55 00:09:12.440 00:09:18.000 Casie Aviles: make some updates on like with the stages. So he added a few, which is, I think

56 00:09:18.170 00:09:19.959 Casie Aviles: makes it clearer, like.

57 00:09:19.960 00:09:20.550 Uttam Kumaran: Cool.

58 00:09:20.890 00:09:26.356 Casie Aviles: Yeah, in progress and pause stuff like that. So based on, you know

59 00:09:27.190 00:09:29.930 Casie Aviles: the work that we’re doing and where it’s at.

60 00:09:30.080 00:09:34.700 Casie Aviles: So yeah, I I guess those are the 2 main oh.

61 00:09:35.050 00:09:40.059 Casie Aviles: channels where I see where I track so slack and notion.

62 00:09:40.730 00:09:44.849 Uttam Kumaran: Okay, what about you, Pius? Like? How’s your what’s your process like.

63 00:09:46.120 00:09:54.880 Payas Parab: I’m very like, I basically do just check the slack of like, what’s the most urgent thing. And then one thing I’m working on now for myself is like at least on some of the

64 00:09:55.030 00:10:20.050 Payas Parab: things that there will be dependencies on other people I’m trying to like set like an arbitrary like I, Ryan, we’re testing it out this weekend right? Or this week, where I was like yesterday, like, Hey, if what’s the day? I need to get you the tickets for all the de work so that we can. I can work on it Thursday night. So I’m like trying to plan a little bit further ahead on like a week sprint. But frankly, it’s similar to you, Thomas. Kind of wake up and figure out what what needs to get done. But I think I’m gonna shift towards

65 00:10:20.260 00:10:27.709 Payas Parab: like, Hey, arbitrary internal deadline for me personally, that I know that I can get this my stuff done by the end of the week, and it’s not

66 00:10:28.230 00:10:36.799 Payas Parab: like I’m handing something off to someone at like some unreasonable hour, and then it like I can’t get to it till the following week, because I have people late. So yeah.

67 00:10:37.800 00:10:42.550 Uttam Kumaran: And in terms of like, literally, the stuff you’re doing. Are you just like you just have like a notepad like.

68 00:10:42.570 00:10:47.769 Uttam Kumaran: are you just like literally just writing down tasks or using any of the client pages themselves.

69 00:10:48.210 00:10:53.959 Payas Parab: I right now? I’m just like writing it out. I I handwrite my days. I’m a little bit old school, so.

70 00:10:53.960 00:10:58.410 Uttam Kumaran: No dude. That’s fine, I mean. Look, this is what I’m this is what I’m saying. I’m just trying to learn, because

71 00:10:58.650 00:11:05.750 Uttam Kumaran: everybody will have their like intraday process, right? But even what you mentioned about like having dependencies and blockers.

72 00:11:05.770 00:11:11.220 Uttam Kumaran: everything that’s everyone’s described here has had a dependency whether it’s on me or client or another person.

73 00:11:11.540 00:11:14.639 Uttam Kumaran: So just trying to find get a gauge for where?

74 00:11:14.780 00:11:19.109 Uttam Kumaran: Cause I feel like I mean, look, we don’t have a process today, and everyone’s getting their shit done. So it’s it’s working.

75 00:11:19.120 00:11:20.170 Uttam Kumaran: I think

76 00:11:20.310 00:11:27.190 Uttam Kumaran: there’s everything working. And then like working, really, really? Well. So I’m just trying to get a sense of what that is, so that that makes sense

77 00:11:31.160 00:11:43.239 Uttam Kumaran: cool. And then, Luke, do you want to go like, what’s your general like to do. List process. Sorry. I know you came a little bit late. We’re just talking a little bit about how everybody tracks their work. So just kind of getting a sense from everybody.

78 00:11:43.560 00:11:56.519 Luke Daque: Yeah. Well, for me, I basically am trying to do us. I am. I’m trying to be hands free as much as possible like I, I use, like, whatever we are currently using, like either notion or

79 00:11:56.710 00:12:02.750 Luke Daque: like the Github issues tracker that we are currently using with Payas.

80 00:12:02.950 00:12:07.160 Luke Daque: So yeah, that’s just me, because I’m trying to

81 00:12:07.360 00:12:10.559 Luke Daque: like for notion. For instance, I’m trying to like at least

82 00:12:11.340 00:12:14.770 Luke Daque: get to know our whole notion. Page.

83 00:12:15.020 00:12:18.299 Luke Daque: I mean whole project, or what whatever you call it.

84 00:12:18.930 00:12:21.160 Luke Daque: Cause? Yeah, there’s a lot in here.

85 00:12:21.440 00:12:24.930 Luke Daque: So yeah, I basically just use that.

86 00:12:28.150 00:12:32.140 Luke Daque: like, whatever tasks are assigned to me and like the priorities.

87 00:12:32.880 00:12:37.419 Luke Daque: and like, if ever we discuss something like in during our calls with

88 00:12:38.037 00:12:45.799 Luke Daque: Nico, or pay us, for example, and as much as possible. I I just write them in in notion, in a task.

89 00:12:45.940 00:12:47.050 Luke Daque: in notion.

90 00:12:48.780 00:12:49.370 Luke Daque: Yeah.

91 00:12:49.370 00:12:54.189 Uttam Kumaran: Is, is like prior like. So let’s talk a little bit, for every task like

92 00:12:54.310 00:12:58.339 Uttam Kumaran: is priority helpful is due date more helpful.

93 00:12:59.049 00:13:04.630 Uttam Kumaran: Is the type of task helpful like. What is the what is the thing that everybody like looks at initially.

94 00:13:08.870 00:13:15.500 Miguel de Veyra: I guess I’ll start this one. But for me it’s basically on who is closest to Demo, to the demo date.

95 00:13:15.500 00:13:16.180 Uttam Kumaran: Okay.

96 00:13:16.880 00:13:17.980 Miguel de Veyra: Because I think you know.

97 00:13:18.080 00:13:22.739 Miguel de Veyra: cause primarily from the AI and Automation side, we’re just working on a lot of Demos.

98 00:13:23.060 00:13:28.910 Miguel de Veyra: So you know, that’s the priority right now and then. Once it’s done, you know. Then the backlog starts coming in.

99 00:13:31.360 00:13:39.970 Uttam Kumaran: And then for pies. And Luke like, Do you guys with Nico? Or you know, with the clients like, does everything have a due date. Is it kind of like.

100 00:13:40.280 00:13:43.100 Uttam Kumaran: just like the next thing kind of piles on?

101 00:13:43.613 00:13:47.860 Uttam Kumaran: Are you? Are you guys setting due dates like, how is it kind of working with you guys today?

102 00:13:50.030 00:13:51.850 Uttam Kumaran: I mean, I think I know the answer. But yeah.

103 00:13:53.340 00:14:01.310 Payas Parab: Yeah, we’ve kind of been going off of like a what’s next type thing? I think I think specifically the client we’ve been working on. It’s a little bit

104 00:14:01.780 00:14:06.339 Payas Parab: they don’t really know what they want. They just kind of want us to do stuff right.

105 00:14:06.340 00:14:06.899 Uttam Kumaran: Yeah, yeah, yeah.

106 00:14:06.900 00:14:12.220 Payas Parab: Is like. So I think, like I mean, that was one of the things. Who, Tom, you know I don’t know. Robert sent you my like.

107 00:14:12.590 00:14:14.629 Payas Parab: Yes, we can improve as a team like.

108 00:14:14.630 00:14:15.040 Uttam Kumaran: Yes.

109 00:14:15.040 00:14:17.709 Payas Parab: Like, what does that end state look like we

110 00:14:18.060 00:14:36.919 Payas Parab: kind of. We sort of had a sense of what the end State looked like, but we didn’t know for sure, right? And like part of it is like, I didn’t know what the end state looked like, so I wasn’t like directing like, Hey, by this date, we’re gonna have this, this and this like, now, I’m starting to get like, okay, how we would work backwards. So I think, like right now, it’s like, and Ryan, correct me. If I’m wrong. It’s kind of just like

111 00:14:37.010 00:14:42.249 Payas Parab: we’re just progressing on certain work streams like we are making great progress across all of them.

112 00:14:42.553 00:14:56.309 Payas Parab: We’re just not working backwards from an end deliverable. But like I think we’ve done a lot of good work for them now. It’s just like a matter of like making sure they see it, which I think both I’m on. And even Jared, which is like crazy, is like starting to have that like Aha! Moment of like.

113 00:14:56.310 00:14:56.790 Uttam Kumaran: Nice.

114 00:14:56.790 00:15:02.899 Payas Parab: I’m seeing the like value. And I, Ryan, I know if you were in the meeting with them on the last meeting with them on like there was.

115 00:15:02.940 00:15:08.099 Payas Parab: It seemed like there was some like like a light bulb going off of like. I see why we’re doing this.

116 00:15:08.100 00:15:08.620 Luke Daque: Yeah.

117 00:15:08.620 00:15:10.949 Uttam Kumaran: Do you think like the solve to that is like.

118 00:15:11.120 00:15:19.700 Uttam Kumaran: for example, like, look in in the teams. I’ve ran from a project management side. I would usually at least start with quarterly goals and then basically make it up

119 00:15:19.840 00:15:21.190 Uttam Kumaran: back into like

120 00:15:21.420 00:15:36.910 Uttam Kumaran: key milestones that are like either on a monthly basis or wherever they land in the month. The nice thing about that is like, you know that, hey? If we deliver a data model, then it’s gonna take another week to set up the dashboard, and you kind of back into the different segments. Is that

121 00:15:36.910 00:15:37.700 Uttam Kumaran: like

122 00:15:37.700 00:16:00.319 Uttam Kumaran: helpful cause I know, for now, like I don’t even know what the key milestones are. But I would. What I’m gonna do is basically try to push, to say, like, what is that high level thing, we need to be able to break it down into tickets and timelines. Because you guys can say, Hey, if we have 5 data models, then that’s gonna each of them gonna take 2 weeks. Then you can kind of back into what’s possible. But if you’re working from just like, here’s the next thing

123 00:16:00.360 00:16:04.769 Uttam Kumaran: you have no idea, like you have just no idea what’s coming up next.

124 00:16:04.770 00:16:05.150 Payas Parab: Hmm.

125 00:16:05.520 00:16:10.830 Uttam Kumaran: So like what would be helpful in terms of like a milestone or a larger milestone perspective.

126 00:16:13.650 00:16:17.070 Payas Parab: Yeah, I think. I think there’s like sort of

127 00:16:17.360 00:16:24.480 Payas Parab: like what that, what the final analytics things look like. And then working backwards and then baking in enough time for like

128 00:16:24.720 00:16:30.219 Payas Parab: data validation and like iterating on it like I think right now, it’s like, Oh, I’m working on

129 00:16:30.460 00:16:46.359 Payas Parab: this view, and like. Therefore we need this like this, like de work stream to build the data model. And then, like as I get it, I’m like, Oh, there’s some things that we just need to like, fix and like do something slightly differently. So there’s like more of an iteration process baked in. So we’d sort of need to like

130 00:16:46.920 00:17:05.859 Payas Parab: work from like, there’s the end state. This is like the data modeling component for like a particular sub project. If that makes sense, like, I consider, like the gorgeous recharge, right? Like optimizing your customer operations. Data like its own work stream right? And like, we sort of just like, did that concurrently alongside, like financial metrics alongside, like

131 00:17:05.859 00:17:23.560 Payas Parab: order, metrics and customer metrics, right and, like all of them, are helpful now and like, because it makes everything come together easily. But if we had like for each work streams and like, Okay, this is the the deliverable. And then like, how do we work backwards and bake in a 2 to 3 week iteration period between client DED.

132 00:17:23.760 00:17:26.170 Payas Parab: Ea like that might have helped a little bit.

133 00:17:26.940 00:17:30.870 Payas Parab: Okay. Well, Ryan, let me know.

134 00:17:30.870 00:17:31.260 Luke Daque: Yeah.

135 00:17:31.260 00:17:33.190 Payas Parab: Or have any opinions on that.

136 00:17:33.190 00:17:45.010 Luke Daque: Yeah, I would. I was gonna say, I agree. And but I I think that’s also something that we need to work on at the moment, because I think that’s where we are also like not very

137 00:17:45.846 00:17:48.340 Luke Daque: I don’t know what’s the term like

138 00:17:49.090 00:17:57.595 Luke Daque: aligned on right like in terms of like, for example, that re recharge thing and gorgeous like.

139 00:17:58.660 00:18:05.760 Luke Daque: I was like creating data models. But so far they were like specific to just one source, data sets.

140 00:18:06.040 00:18:12.420 Luke Daque: or like, just for gorgeous or just for recharge. But I wasn’t really working on

141 00:18:12.890 00:18:14.879 Luke Daque: like combining both of them.

142 00:18:15.360 00:18:17.279 Payas Parab: And that’s that’s and that even

143 00:18:17.280 00:18:21.140 Payas Parab: like from a mon like, we got that feedback last time, right where I was like, Okay, cool.

144 00:18:21.140 00:18:21.690 Payas Parab: Feel good about

145 00:18:21.690 00:18:27.080 Payas Parab: the gorgeous data. We feel good about how we’re going to visualize it, and then we can start to iterate with Justin. And

146 00:18:27.890 00:18:30.979 Payas Parab: but then he’s like, well, recharge is a big part of this, and we’re like.

147 00:18:31.050 00:18:34.299 Payas Parab: that’s the 1st we’re hearing of that right and and like

148 00:18:34.310 00:18:44.030 Payas Parab: in in like. And that’s totally not on Ryan. That’s on me, right is like I could have somehow been like, here’s the end state we’re going towards, and then gotten that feedback from Aman earlier, that, like.

149 00:18:44.030 00:18:53.100 Uttam Kumaran: But that’s I don’t know. I feel like part of that is also on. It’s also on Nico, like, I think we should see ahead and basically back in. And this is where we play defense

150 00:18:53.120 00:19:00.190 Uttam Kumaran: like, look if they’re like, Hey, we want gorgeous data, then we basically have to say, here is the current roadmap. What would you like us to replace

151 00:19:01.200 00:19:07.550 Uttam Kumaran: that with? Right? If you don’t have the roadmap? Then everything is just top priority added to the top of the thing.

152 00:19:07.670 00:19:17.270 Uttam Kumaran: So this is where also, like, even in that like, Hey, something came up, or we want to add requirements. Fine. Yeah, maybe we could have like seen a little bit ahead. But frankly, it should be like, what do we need to swap out?

153 00:19:17.895 00:19:25.510 Uttam Kumaran: If everything, if we’re going week by week. Then there’s something to swap out, because this is like added to the next thing. But that’s a hectic life, you know. So

154 00:19:25.570 00:19:31.350 Uttam Kumaran: that’s kind of where I’m kind of seeing. This is like, okay, I think there is a clear improvement that we can do, which is like, can we get?

155 00:19:31.370 00:19:33.819 Uttam Kumaran: Can we start by getting a 1 month ahead

156 00:19:33.930 00:19:41.780 Uttam Kumaran: right in terms of tasks, because weeks go by fast for us? Can we start by getting one month ahead. Then can we start getting 3 months ahead?

157 00:19:42.130 00:19:57.220 Uttam Kumaran: Can we? On this project? It’s nice because we have a key dashboard deliverables, sources, data models. And that way when Nico is like cool, I have. This is the high level goal. Can Pius and Ryan. Can you guys break this down into the key

158 00:19:57.290 00:20:01.509 Uttam Kumaran: tasks and then estimate on how long it’s gonna take.

159 00:20:01.590 00:20:04.280 Uttam Kumaran: Then he has all the information he needs to track.

160 00:20:04.500 00:20:08.879 Uttam Kumaran: So I think that’s what I’m gonna kind of push a little bit on and try to see whether we can make that happen.

161 00:20:08.880 00:20:27.089 Payas Parab: Yeah, I I will say, like, it’s not fully because I think like the way I look at like me, Nico and Ryan’s role, like, I think, getting some of the requirements from client, especially when you think about the analytics side like, frankly, that’s on me like you know what I mean, like, I think it’s like Nico is helping to structure the project. But like actually getting the requirements, it’s like.

162 00:20:27.220 00:20:44.129 Payas Parab: you know, it’s like we knew we were working towards a customer dashboard, and I think, like in hindsight, I would have sat down and been like, Hey, it’s worth the 2 h, 3 h, even if I’m not like actually making any visualizations. To like be like. This is what that will look like, right like. This is what that dashboard will look like. And then like giving it to Nico.

163 00:20:44.130 00:20:44.750 Payas Parab: I see.

164 00:20:44.750 00:20:57.439 Payas Parab: Fill in the gaps right like like that’s also I want to be clear like we. We just need to. Just because because also to be fair like this project was my 1st time working with you guys. So it’s like, it’s not like anyone was. You know what I mean like, and I’m like, if I was looking at it, and I was giving my.

165 00:20:57.440 00:20:57.760 Uttam Kumaran: Totally.

166 00:20:57.760 00:21:14.780 Payas Parab: Feedback. I would be like, well, pious, you know, like I wouldn’t be like, hey? Nico was supposed to figure that out. It’s like, well, pious is creating the end layer pious needs to figure out what that looks like. Then Nico needs to ingest those and requirements and then break those out into reasonable tasks for Ryan right and and like working backwards. And so I think

167 00:21:15.000 00:21:27.330 Payas Parab: again, like like Robert’s, done a great job historically, of like getting the clients an end state right? Like, if you saw the original Javi, Doc, it’s like, this is what we’re going to deliver these the business questions. You’re going to be able to answer like I never did that for like

168 00:21:27.380 00:21:35.750 Payas Parab: customer service and lifetime value, I never did that, for, like financial metrics, right? I just sort of started doing it. So I’m like looking at myself, and I would be like I should have

169 00:21:35.920 00:21:41.409 Payas Parab: done something more like Robert like in that, and then handed that to Nico to help us structure the project.

170 00:21:42.370 00:21:51.699 Luke Daque: That actually makes sense. And yeah, considering, like, we’re pretty new, as like a team like working together. This is this, yeah, this is a good.

171 00:21:52.000 00:21:54.477 Luke Daque: some sort of retrospective right, like

172 00:21:55.106 00:21:55.540 Payas Parab: Oh, okay.

173 00:21:55.540 00:21:58.590 Luke Daque: You know, to to be able to discuss, like.

174 00:21:58.590 00:21:59.220 Payas Parab: Talking about.

175 00:21:59.220 00:22:03.050 Luke Daque: How we can improve things, and maybe next time, when we do

176 00:22:03.060 00:22:09.619 Luke Daque: like our meetings with Nico, for example, by us for the Javi off internal meetings. For example, then.

177 00:22:09.620 00:22:10.330 Luke Daque: yeah, we can.

178 00:22:10.330 00:22:13.219 Luke Daque: We can discuss about that like.

179 00:22:13.630 00:22:25.889 Payas Parab: Yeah, exactly. And I’m comfortable, like Nico, holding me accountable to be like, hey, I’m planning the next month right? And getting us organized around the client. You have to give us what that end state is, and it’s like, if no one holds me accountable to that, I’m just gonna be like cool.

180 00:22:25.890 00:22:26.490 Uttam Kumaran: Totally.

181 00:22:26.490 00:22:29.240 Payas Parab: Solve this problem for the meeting I’m on like. And again.

182 00:22:29.240 00:22:29.600 Uttam Kumaran: Totally.

183 00:22:29.600 00:22:42.479 Payas Parab: In in future. I’m hoping not to do that, but that’s just what it’s been right. And so if Nico and Ryan want to hold me accountable like. You have to tell us what that end state is, and maybe you have to go to the client for that. And we, you know that point. We need Robert. And you as well, right?

184 00:22:42.480 00:22:43.170 Payas Parab: Yeah, yeah. Yeah.

185 00:22:43.170 00:22:49.079 Payas Parab: Client service experience. Like, there’s sort of that chain that has to like, go on and I think

186 00:22:49.410 00:23:00.290 Payas Parab: it can start with like us, just having like a joint structure of like I don’t know. We’re planning the next few weeks, and it doesn’t have to be like, you know, this week we’re gonna do this this week. We’re gonna do this and like super over the top plan.

187 00:23:00.850 00:23:14.689 Payas Parab: Like fancy Jira board of like all that shit like just more like, hey? This is roughly the idea for next month, and like, if you hold me accountable like, hey? Pious! You need to figure out what we are going to give to the client at the end of the month, and then I can look at it and be like, Okay, well.

188 00:23:14.820 00:23:24.540 Payas Parab: okay, cool. We need to like expectation set with the client which I feel like always. Robert has been better out than me. Then I can be like, Hey, Robert, you need to expectation. Set this, or like, help me guide this and scope this out.

189 00:23:24.680 00:23:47.860 Payas Parab: Then we get it to the client. Then we’re like more structured. And then we have, like a week of like a 1st stab at it 2 to 3 weeks of iteration, and then like a 4th week on that, and it can be like at like a sub project level, right? So like customer service could be one sub project and like again, it doesn’t need a super fancy tracker like I wouldn’t mind if it’s just like one document that’s like. Here’s the 8 projects, and here’s where we’re at on each of them. And

190 00:23:47.960 00:23:54.459 Payas Parab: you know, to whatever extent anyone wants to hold me accountable to like getting what that end state looks like. I’m I’m open to that.

191 00:23:55.050 00:24:04.079 Uttam Kumaran: So you’re totally right in that. There’s there’s 2 things there’s 1 is like planning the projects out themselves, making sure that if a project is like a around like a key

192 00:24:04.140 00:24:28.900 Uttam Kumaran: business unit, or it’s around a set of dashboards that all those we have requirements for. And then then there’s the day to day of just like, what’s on the agenda today, like, what’s on the agenda this week? Definitely, I think we have been doing a lot more of the latter, and not enough of the former, where we spend a lot of the time, the the only time we have with the client every week on what’s being delivered. But there isn’t like a work stream or tasks around.

193 00:24:29.203 00:24:33.309 Uttam Kumaran: Getting those requirements right? Because you spending time on project planning is a task.

194 00:24:33.360 00:24:34.679 Uttam Kumaran: And so even that

195 00:24:34.960 00:24:47.419 Uttam Kumaran: I wanna make sure it gets tracked somewhere. And then also again, we as long as we have all the projects basically set. Then we can say, Hey, this project is in planning until, like me and Robert have signed off, the client is signed off.

196 00:24:47.450 00:24:49.990 Uttam Kumaran: and I also signed off on like, what’s

197 00:24:50.030 00:25:15.959 Uttam Kumaran: what is in there? Then it can move on to like kind of breaking it up into tickets. So I’ll think of something there. Because basically look, we have, there’s gonna be 2 things. There’s 1 is like the day to day. The second thing is like cool. Let’s look to see if we can get a month ahead, and then on the sales side is when we could start talking about like account management, higher level retention. Then then the nice thing is on a month level or on a quarterly level. I can show all the work that we’ve done for them, and then say.

198 00:25:15.990 00:25:21.720 Uttam Kumaran: hey? You may need to consider putting more hours to this. You want us to get more done, or is there other stuff you want us to take?

199 00:25:21.980 00:25:24.289 Uttam Kumaran: So this all layers on each other.

200 00:25:24.720 00:25:30.300 Uttam Kumaran: So I’m gonna kind of convey this, and we’ll kind of get some stuff going on this I know for the AI

201 00:25:30.400 00:25:35.900 Uttam Kumaran: folks. This is more like we’re kind of like a little bit on just building this service arm in general.

202 00:25:35.910 00:25:51.348 Uttam Kumaran: But the nice thing is, we will avoid some of these issues. When we start doing work, and that when we take on work for clients, we will bake in some time specifically for this you know, for this planning. So that’s the biggest thing that

203 00:25:51.880 00:25:56.100 Uttam Kumaran: I wanna work on is we have this, this planning that goes on stuff for a given month.

204 00:25:56.110 00:25:58.479 Uttam Kumaran: We know everything that’s being worked on. And then

205 00:25:58.850 00:26:05.850 Uttam Kumaran: everybody on this call comes to work and knows exactly what they need to do. And we’re not all stressed by like what’s coming up next or not knowing.

206 00:26:06.990 00:26:08.499 Uttam Kumaran: You know. And and again.

207 00:26:08.510 00:26:13.469 Uttam Kumaran: I we. Our goal is always to get things on fast, and be the fastest to do them.

208 00:26:13.590 00:26:19.429 Uttam Kumaran: But I will say there’s a benefit to doing things slower, not only for, like our sanity, but also

209 00:26:19.800 00:26:35.889 Uttam Kumaran: the timeline. The client may not care whether it’s a week or 2 weeks. If we say a week, they’re gonna be happy for, say, 2 weeks they may still be happy, right? And so that’s what we need to figure out. And what my next topic is gonna be, maybe I can start with the AI guys, I want to start thinking about what are the different task types?

210 00:26:37.362 00:26:47.950 Uttam Kumaran: or maybe actually, maybe I’ll start with pious first, st because I know you have to go so you can kind of maybe brain dump. And then we can talk with the AI folks like, what are the types of tasks or requests

211 00:26:47.960 00:26:50.140 Uttam Kumaran: that you get on the data side.

212 00:26:50.170 00:26:52.600 Uttam Kumaran: And basically, how can we standardize

213 00:26:52.740 00:26:58.249 Uttam Kumaran: those requests? Meaning like just brain dump, like, what are the types of asks we get

214 00:26:58.430 00:27:02.479 Uttam Kumaran: whether it’s like this, dashboard’s broken, whether it’s like I need a new metric.

215 00:27:02.933 00:27:07.000 Uttam Kumaran: And then let’s just list them all down. And then I’m gonna turn these into things that we can track.

216 00:27:07.900 00:27:19.799 Payas Parab: Yeah, I think just to list a few. I mean, one of them is like, it’s like a new new feature, right? I think, like a new new type of view that they don’t currently have in the visualization. Right is like, Hey, can we cut it?

217 00:27:20.536 00:27:33.100 Payas Parab: By this? And it may be a feature that we don’t actually have built in right? And like the data model where we haven’t had it flow through for whatever like, you know, like we’ve we that wasn’t in the original scope. So I think there’s like this new new feature

218 00:27:33.170 00:27:34.840 Payas Parab: work stream. That kind of

219 00:27:34.860 00:27:49.840 Payas Parab: relates to engineering and analytics where it’s like, Hey, I want to see the data by this. We go click in and don’t see it there. The the other one is. Something’s broken. Obviously, right? That’s the easy easy one. Typically, the breakage is something simple. Then there’s like

220 00:27:50.030 00:27:59.759 Payas Parab: the other big ask, I think, is like around like data strategy questions essentially like, Hey, you guys, are the expert on this like, Tell us the best way to set this up.

221 00:27:59.890 00:28:02.750 Payas Parab: And sometimes the best way to set that up is

222 00:28:02.870 00:28:05.929 Payas Parab: like it. It involves, like the engineering side as well. Right.

223 00:28:05.930 00:28:07.889 Uttam Kumaran: What does setup mean? So what up.

224 00:28:08.370 00:28:10.550 Payas Parab: I’ll give an example is like the 5 tran right like.

225 00:28:10.550 00:28:11.060 Uttam Kumaran: Oh, okay.

226 00:28:11.060 00:28:24.450 Payas Parab: Like, like the 5 trend Google sheets type setup thing. It was like, Okay, cool. We want to do this. And I want to use Google sheets, because, like, that’s where all my Ops guys use. And so we have to kind of solve that problem. So it’s like, there’s sort of like a data

227 00:28:24.520 00:28:28.399 Payas Parab: or tooling strategy question that ends up being more involved than like

228 00:28:28.530 00:28:31.769 Payas Parab: we have a off the top of our head. Answer for you, you know.

229 00:28:33.150 00:28:34.120 Uttam Kumaran: Okay, makes sense.

230 00:28:34.120 00:28:36.300 Payas Parab: Those would be the big categories that come to mind.

231 00:28:36.300 00:28:41.280 Uttam Kumaran: Okay, Ryan, you want to go on the data side.

232 00:28:43.260 00:28:46.070 Luke Daque: Yeah, I think I agree with Payas as well.

233 00:28:49.870 00:28:58.679 Luke Daque: yeah, I I yeah. I’m also like quite thinking, what else I can add to that. Maybe we can go with the AI first.st

234 00:29:00.970 00:29:05.330 Uttam Kumaran: Okay, yeah. Miguel, or maybe yeah, maybe, Casey, do you want to go?

235 00:29:05.470 00:29:11.209 Uttam Kumaran: Just talk a little bit about like, what are the key tasks that you’ve seen now that you’ve been working with us for a while.

236 00:29:11.895 00:29:14.279 Uttam Kumaran: If you could just break them down into categories.

237 00:29:16.690 00:29:20.580 Casie Aviles: Yeah, sure. For the tasks, I think.

238 00:29:20.670 00:29:28.259 Casie Aviles: Mostly I’ve been doing more that that’s focused on like like the clients. So

239 00:29:28.670 00:29:34.473 Casie Aviles: in the Kanban board like I would check the the demo, or, for example,

240 00:29:35.540 00:29:39.110 Casie Aviles: for sasaholic, we have a sasholic demo, and then

241 00:29:39.460 00:29:42.720 Casie Aviles: the other one is so it’s more client centric.

242 00:29:45.340 00:29:50.200 Casie Aviles: I guess that’s 1 category there are also, like some

243 00:29:50.850 00:29:57.576 Casie Aviles: some tasks peppered in that are more, I’d say atomic, or like, they’re just, you know,

244 00:29:59.265 00:30:07.969 Casie Aviles: simpler tasks, like, you know, quick adjustments to the prompt like doing like quick scraping tasks and stuff.

245 00:30:09.221 00:30:14.350 Casie Aviles: I guess. Yeah. And then there are some internal tasks as well like, for example,

246 00:30:15.150 00:30:22.120 Casie Aviles: the the agent, like I categorize them into like a a zoom agent, and then another would be for

247 00:30:23.320 00:30:26.560 Casie Aviles: for the lead research agent. So I guess

248 00:30:27.810 00:30:31.029 Casie Aviles: that’s those are the kinds of tasks that I work on.

249 00:30:31.870 00:30:38.570 Uttam Kumaran: There’s almost like new agent development, new tool development, prompt adjustments, scraping tasks.

250 00:30:39.700 00:30:40.820 Uttam Kumaran: Yeah, go ahead, Miguel.

251 00:30:44.040 00:30:48.690 Miguel de Veyra: Oh, let me think!

252 00:30:53.770 00:30:59.999 Uttam Kumaran: Cause. So let’s let’s talk about agent modifications like, what are some of the if, if our like okay, cool, we have an existing agent.

253 00:31:00.100 00:31:06.430 Uttam Kumaran: And we want to make changes like, what are some of the changes that are typically being made like prompt adjustments?

254 00:31:06.560 00:31:08.089 Uttam Kumaran: We’re like adding tools.

255 00:31:08.410 00:31:09.230 Miguel de Veyra: Yeah.

256 00:31:09.230 00:31:10.919 Uttam Kumaran: Knowledge base updates.

257 00:31:16.730 00:31:19.230 Uttam Kumaran: You’re updating the entire workflow.

258 00:31:20.040 00:31:21.910 Miguel de Veyra: And then, yeah, the documentation.

259 00:31:22.680 00:31:24.869 Uttam Kumaran: Yeah. And documentation is huge.

260 00:31:25.680 00:31:29.530 Uttam Kumaran: It was something we’ll we were not gonna talk about today. But we’ll talk about in a later meeting.

261 00:31:30.625 00:31:31.600 Uttam Kumaran: Because

262 00:31:32.400 00:31:39.220 Uttam Kumaran: there’s nothing to document until we have all of our like until I until we can make sure that our normal works getting done. I don’t want to put documentation on everyone’s plate.

263 00:31:39.220 00:31:39.555 Miguel de Veyra: Yeah.

264 00:31:39.890 00:31:41.732 Uttam Kumaran: But it’s something that we will

265 00:31:42.350 00:31:43.500 Miguel de Veyra: Standardized Offer.

266 00:31:43.500 00:31:46.730 Uttam Kumaran: Will standardize. Okay.

267 00:31:51.170 00:31:57.069 Uttam Kumaran: yeah, bias. I saw your thing. So bug feature end, new data sources. Yeah.

268 00:31:57.515 00:32:00.440 Uttam Kumaran: yeah. And as part of Github issues. Yeah. So

269 00:32:00.790 00:32:07.389 Uttam Kumaran: I think we’re gonna I’m gonna make some decisions on how we’re gonna use Github issues versus notion and how to get that all working, but

270 00:32:07.410 00:32:11.479 Uttam Kumaran: good to make sure. I mean, this is all pretty. This is all lines up with what I’m thinking as well.

271 00:32:13.160 00:32:17.859 Uttam Kumaran: again for me. I’m I’m very familiar on how to kind of like establish these on the data side. But

272 00:32:18.100 00:32:22.310 Uttam Kumaran: on the AI side, just wanna make sure this all lines up. And then also, as an engineering org.

273 00:32:22.360 00:32:24.789 Uttam Kumaran: there’s some things that will line up between us.

274 00:32:25.449 00:32:28.939 Uttam Kumaran: That that will be really nice things like documentation, and

275 00:32:28.980 00:32:33.209 Uttam Kumaran: we’ll find ways to add in time, or like tech debt.

276 00:32:33.250 00:32:36.120 Uttam Kumaran: weeks or stuff where we can go clean. Stuff up so

277 00:32:40.400 00:32:43.520 Uttam Kumaran: cool. Alright, thanks bye, son, you have to leave, so feel free.

278 00:32:44.950 00:32:47.389 Payas Parab: Always pleasure chatting with you guys. Alright.

279 00:32:47.390 00:32:47.970 Uttam Kumaran: Talk to you soon.

280 00:32:48.110 00:32:49.550 Miguel de Veyra: Take care. See you later.

281 00:32:49.550 00:32:50.280 Luke Daque: Thanks.

282 00:32:51.780 00:32:55.389 Uttam Kumaran: Cool, so that makes sense anything else to add, Luke.

283 00:32:56.466 00:33:00.309 Luke Daque: For categories. I think we still we we have.

284 00:33:00.580 00:33:05.820 Luke Daque: We also have like data modeling and data visualization as well.

285 00:33:08.420 00:33:09.910 Luke Daque: Yeah. But it’s, I think.

286 00:33:10.340 00:33:14.590 Luke Daque: yeah, essentially, that’s what we are doing. And maybe data ingestion.

287 00:33:14.800 00:33:21.460 Luke Daque: Well, that would be either like 5 tran or whatever 3rd party tool we are using, or if we need to do it

288 00:33:22.060 00:33:23.959 Luke Daque: ad hoc, like on our own

289 00:33:24.568 00:33:28.740 Luke Daque: like, how we used to be doing it with Patrick and the others.

290 00:33:28.740 00:33:29.160 Luke Daque: Yeah.

291 00:33:29.160 00:33:35.279 Luke Daque: Or in case, you know, in case there’s no 3rd party provider for the data ingestion.

292 00:33:36.930 00:33:42.800 Luke Daque: what else any other categories that I can think of but but so far that’s like the main

293 00:33:42.910 00:33:45.730 Luke Daque: main things that we’ve been doing.

294 00:33:48.790 00:33:51.499 Luke Daque: Yeah. And as a going.

295 00:33:51.640 00:34:00.180 Luke Daque: I think, with in terms of like Github issues and notion. I think there’s a way we can connect. Because I was. I can. I am able to connect

296 00:34:00.890 00:34:04.459 Luke Daque: Prs and issues to a notion document. So

297 00:34:04.985 00:34:09.769 Luke Daque: yeah, maybe we can find a way to automate that I haven’t done the research. But I think

298 00:34:09.790 00:34:11.040 Luke Daque: that’s possible.

299 00:34:11.360 00:34:12.030 Uttam Kumaran: Okay.

300 00:34:12.320 00:34:18.870 Luke Daque: Like anytime an issue gets created in a repository. Maybe it can automatically create a notion

301 00:34:18.989 00:34:22.250 Luke Daque: task and stuff like that. So that would be great.

302 00:34:22.600 00:34:29.040 Luke Daque: But yeah, currently, I’m just doing it manually, but it’s fine as well, only takes like 5 seconds or something

303 00:34:29.219 00:34:31.200 Luke Daque: to add it to the notion document.

304 00:34:32.050 00:34:32.639 Uttam Kumaran: Okay.

305 00:34:35.739 00:34:37.570 Uttam Kumaran: great. And so, okay, we have.

306 00:34:38.139 00:34:40.150 Uttam Kumaran: This is all really, really helpful.

307 00:34:40.429 00:34:44.074 Uttam Kumaran: Another thing I want to talk a little bit about is

308 00:34:46.199 00:34:51.570 Uttam Kumaran: like agent documentation and tool documentation like for Miguel and Casey, like

309 00:34:51.580 00:34:56.069 Uttam Kumaran: we’re now building with several agents. We have stuff that are. That’s external. We have stuff

310 00:34:56.110 00:35:06.099 Uttam Kumaran: that’s internal. I want to think through like what a good documentation structure could be. Also, it’s like, even today, like Casey, I just pinged you.

311 00:35:07.470 00:35:08.860 Uttam Kumaran: I just pinged you like.

312 00:35:08.980 00:35:14.030 Uttam Kumaran: can I make a change to one of the lead research?

313 00:35:16.360 00:35:17.100 Uttam Kumaran: Like

314 00:35:17.440 00:35:23.589 Uttam Kumaran: prompts? But for me, it’s like, Okay, where do I go to find that? Is it? Is it? Somewhere where it’s version controlled

315 00:35:23.610 00:35:27.630 Uttam Kumaran: things like that. So kind of want to think on the agent side, like what?

316 00:35:27.790 00:35:30.090 Uttam Kumaran: What the documentation we actually

317 00:35:30.651 00:35:36.940 Uttam Kumaran: like, what the documentation needs to have, also what the request format for new agents needs to look like.

318 00:35:37.379 00:35:39.269 Uttam Kumaran: I don’t know if you guys have any thoughts there.

319 00:35:41.303 00:35:52.090 Miguel de Veyra: Yeah, actually, that was one of the few things we were when we were in the last engineering sync. Because there’s really like, no way we can have, like some sort of version control, like like Github, right.

320 00:35:53.010 00:35:55.449 Uttam Kumaran: But why can’t? We can’t connect an 8 N. To Github.

321 00:35:57.370 00:36:02.369 Miguel de Veyra: No, I mean when when we’re building agents like, for example, if you want to change a prompt or something like that.

322 00:36:03.640 00:36:05.570 Uttam Kumaran: But that’s what I’m saying, I said. Why.

323 00:36:05.620 00:36:10.230 Uttam Kumaran: I’m pretty sure you can save workflows into Github.

324 00:36:11.706 00:36:17.379 Miguel de Veyra: No. I meant like, for example, we want to change something for us as a holiday like the agent. Right? We want to update there.

325 00:36:18.290 00:36:22.870 Miguel de Veyra: Are you saying we connect that to get up every time we like change the prompt.

326 00:36:24.470 00:36:29.340 Uttam Kumaran: I guess what I’m saying is that we need to have some way to be able to collaborate

327 00:36:29.530 00:36:31.980 Uttam Kumaran: effectively on these agents.

328 00:36:31.980 00:36:32.320 Miguel de Veyra: Yeah.

329 00:36:32.621 00:36:39.560 Uttam Kumaran: And like, have version control and have reviews and stuff like that like this is just gonna have to become a lot less

330 00:36:40.040 00:36:41.120 Uttam Kumaran: janky.

331 00:36:43.590 00:36:46.960 Uttam Kumaran: So one is like for relevance and stuff, sure. But

332 00:36:47.600 00:36:50.569 Uttam Kumaran: again, like I don’t. I don’t know how like.

333 00:36:50.650 00:36:53.496 Uttam Kumaran: It’s tough, because we can’t. We’re not. I’m not gonna be able to trust

334 00:36:53.880 00:37:00.570 Uttam Kumaran: to do anything production, grade and relevance. If I can’t see the code like we can’t, we can’t do larger scale testing

335 00:37:01.132 00:37:07.580 Uttam Kumaran: because we’re gonna get to the point where, when we push agents, they run through testing automatically, we do testing.

336 00:37:07.630 00:37:15.470 Uttam Kumaran: So that’s why I’m like trying to think through what the best way to document, for example, like right now, I can’t go to the lead researcher and see like

337 00:37:16.250 00:37:19.409 Uttam Kumaran: I can’t make a change and then measure the impact. And then like.

338 00:37:19.730 00:37:20.950 Uttam Kumaran: do you know what I mean?

339 00:37:22.493 00:37:24.729 Uttam Kumaran: So I don’t know. Like, what do you guys think.

340 00:37:27.470 00:37:30.750 Miguel de Veyra: One of the things we could do.

341 00:37:31.760 00:37:36.400 Miguel de Veyra: I guess it’s though it’s kind of it’s very. It’s gonna be very manual.

342 00:37:36.680 00:37:45.280 Miguel de Veyra: is, for example, we create like a doc, an Asian documentation that’s just use as a holic for now and then we put the prompt there. So it’s saved somewhere.

343 00:37:45.900 00:37:49.700 Miguel de Veyra: And then, but yeah, how do we track.

344 00:37:49.700 00:37:52.059 Uttam Kumaran: Yeah, I mean, you’ll have to copy. Paste it every time.

345 00:37:52.290 00:37:53.340 Miguel de Veyra: Yeah, yeah.

346 00:37:53.970 00:37:55.799 Uttam Kumaran: Does relevance have an Api.

347 00:37:57.990 00:37:59.670 Miguel de Veyra: Yeah, yeah, they do. They do.

348 00:38:01.950 00:38:04.939 Uttam Kumaran: Are you able to get things like prompts and stuff.

349 00:38:06.230 00:38:08.080 Miguel de Veyra: I believe they have like this.

350 00:38:08.780 00:38:12.319 Miguel de Veyra: Yeah, they have the Api to create an agent with. Let me double check it.

351 00:38:20.230 00:38:21.860 Miguel de Veyra: Api.

352 00:38:22.510 00:38:25.280 Uttam Kumaran: It looks like they have create agent, but they don’t have like.

353 00:38:25.670 00:38:26.580 Miguel de Veyra: Update.

354 00:38:27.880 00:38:29.030 Miguel de Veyra: Well, let me show you.

355 00:38:30.440 00:38:32.320 Uttam Kumaran: They don’t have like getting information.

356 00:38:39.680 00:38:41.640 Miguel de Veyra: Yeah. I don’t think they do like.

357 00:38:41.910 00:38:43.730 Uttam Kumaran: They just have list conversations.

358 00:38:43.730 00:38:47.549 Miguel de Veyra: Yeah, yeah, they have list. And then, yeah, conversation

359 00:38:47.980 00:38:50.840 Miguel de Veyra: create an agent. I don’t think this is an Api.

360 00:38:51.550 00:38:52.859 Miguel de Veyra: Yeah, it’s not.

361 00:38:54.080 00:38:54.730 Uttam Kumaran: Okay.

362 00:38:54.970 00:38:56.480 Miguel de Veyra: So it’s a lot of copy paste.

363 00:38:58.820 00:39:00.569 Uttam Kumaran: Yeah. It’s been a tough.

364 00:39:03.200 00:39:09.359 Miguel de Veyra: I guess that just boils down to discipline, then that you know we have to update that documentation.

365 00:39:11.800 00:39:12.570 Uttam Kumaran: Yeah.

366 00:39:13.170 00:39:21.380 Miguel de Veyra: And then one of the things we built before utham was basically like an automated tester of an agent.

367 00:39:22.920 00:39:24.060 Uttam Kumaran: Okay.

368 00:39:26.570 00:39:31.789 Miguel de Veyra: let me check if I have the code for it. But it was using open AI. But I guess you know.

369 00:39:31.920 00:39:35.900 Miguel de Veyra: basically, it’s just another relevance agent like Qa. Tester. Right?

370 00:39:36.880 00:39:41.539 Uttam Kumaran: This is where I wanted to run like. This is where I was. I was hoping we could have something like trace loop.

371 00:39:41.660 00:39:44.240 Uttam Kumaran: or something that we run with every single agent.

372 00:39:44.870 00:39:47.760 Uttam Kumaran: Because I want to have evals for every agent.

373 00:39:47.820 00:39:56.720 Uttam Kumaran: Basically because whenever we get a request, we’re gonna have test cases. And then we’re gonna have. We’re gonna have outputs. And then we can measure like how close our agents are to the output.

374 00:39:56.970 00:39:57.940 Miguel de Veyra: Yeah, right?

375 00:39:58.530 00:40:01.259 Miguel de Veyra: I think relevance has that feature to be honest.

376 00:40:01.940 00:40:05.339 Miguel de Veyra: But we have. I haven’t really explored it. To be honest, because

377 00:40:07.010 00:40:15.600 Miguel de Veyra: I think they do. I’ll explore that. But I think there’s an option there where basically advanced settings.

378 00:40:16.050 00:40:17.929 Miguel de Veyra: Now it’s just naming tasks.

379 00:40:20.520 00:40:23.400 Uttam Kumaran: And for relevance are we using open? Are we using azure.

380 00:40:24.650 00:40:28.709 Miguel de Veyra: Yes, we just migrated the most used agent side now.

381 00:40:28.720 00:40:30.710 Uttam Kumaran: Like earlier when we were on the call.

382 00:40:30.920 00:40:34.700 Miguel de Veyra: Actually didn’t have it like a few days ago. But now they do. So

383 00:40:34.820 00:40:36.830 Miguel de Veyra: we just maybe editions of site.

384 00:40:40.160 00:40:43.860 Miguel de Veyra: because we actually ran out of credits. Now, Casey, for Openai.

385 00:40:43.860 00:40:45.100 Casie Aviles: Yeah, yeah.

386 00:40:47.860 00:40:53.050 Uttam Kumaran: Need to. So for the AI just need to think about how to version control and.

387 00:40:53.050 00:40:53.620 Miguel de Veyra: Control.

388 00:40:53.620 00:40:56.390 Uttam Kumaran: Of discipline, on, on relevance.

389 00:40:59.110 00:41:06.179 Miguel de Veyra: I think what even we could do with them is create like, for example, for Sasaholic, we have, like a staging agent.

390 00:41:06.200 00:41:09.709 Miguel de Veyra: and then, like a production agent. I think that’s the best way to do it.

391 00:41:16.270 00:41:21.180 Uttam Kumaran: It’s also more like dude. I wanna make like a couple of changes to the prompt and just test it.

392 00:41:24.890 00:41:25.720 Uttam Kumaran: But

393 00:41:26.220 00:41:31.279 Uttam Kumaran: yeah, okay, we’ll we’ll have to think longer about this. I’ll have to call some friends and ask them how they’re doing this.

394 00:41:35.830 00:41:39.590 Uttam Kumaran: And then I dude, I guys, I really want to think about something for evals.

395 00:41:40.990 00:41:42.240 Miguel de Veyra: Yeah, I just noticed.

396 00:41:42.240 00:41:50.789 Uttam Kumaran: So for Sasaholic, we have, we have 10 examples and 10 example outputs. Right? So when we make a change, I want to make sure that those don’t break

397 00:41:52.670 00:41:55.720 Uttam Kumaran: alright. So whether it’s trace loop, whether it’s something else like

398 00:41:56.230 00:41:58.070 Uttam Kumaran: really want us to pick something.

399 00:41:58.800 00:41:59.550 Miguel de Veyra: Yeah.

400 00:41:59.730 00:42:05.099 Uttam Kumaran: The thing is, I don’t know. I guess you can’t run relevance with Traceloop.

401 00:42:05.130 00:42:11.399 Uttam Kumaran: but I care less about trace loop. Understanding each of the calls. I just care about the the Evals.

402 00:42:13.010 00:42:20.669 Casie Aviles: Yeah, because currently, right now, it’s working, just, you know, tracing the calls and the tokens and the costs. That’s how it’s working right now.

403 00:42:21.210 00:42:28.870 Uttam Kumaran: But that I guess it’s like less of like a priority. Biggest thing is like we can get the Evals.

404 00:42:29.935 00:42:40.039 Miguel de Veyra: We, I built something like that before with them, we’re basically it’s a bit expensive, though, to be honest, because it’s in relevance. But basically, what we did was we take

405 00:42:40.140 00:42:48.900 Miguel de Veyra: the entire conversation history. We we have like an agent that processes it and then grades it. And then we store that to air table like that was our Eval before

406 00:42:49.430 00:42:51.280 Miguel de Veyra: it was very manual. But you know

407 00:42:51.640 00:42:53.670 Miguel de Veyra: I mean manual in terms of setting it up.

408 00:42:54.860 00:42:55.830 Miguel de Veyra: I think we could do something.

409 00:42:55.830 00:43:01.729 Uttam Kumaran: There. There are some. There are some like things that are available open source. Now that I think

410 00:43:02.860 00:43:05.330 Uttam Kumaran: you may find to be better.

411 00:43:07.090 00:43:09.259 Uttam Kumaran: Let me send you a couple of things

412 00:43:14.070 00:43:18.829 Uttam Kumaran: like we got we should run something like this that arise has

413 00:43:20.540 00:43:22.399 Uttam Kumaran: I’m gonna send in the AI team.

414 00:43:30.510 00:43:34.119 Uttam Kumaran: These guys, I heard are the best, probably open source observability.

415 00:43:35.009 00:43:37.879 Uttam Kumaran: But I think we should try something like arise.

416 00:43:38.150 00:43:41.489 Uttam Kumaran: and we could like, I’m happy to sign up for them if we want to.

417 00:43:41.790 00:43:45.699 Uttam Kumaran: They have like great evaluations, basically.

418 00:43:47.810 00:43:51.710 Uttam Kumaran: But again, like we may have, we may not be able to use relevance.

419 00:43:52.110 00:43:52.480 Miguel de Veyra: You know.

420 00:43:52.480 00:43:53.420 Uttam Kumaran: Some of these.

421 00:43:55.150 00:43:55.530 Casie Aviles: Yeah.

422 00:43:55.530 00:44:05.400 Miguel de Veyra: Actually, there is a way we can do this looking at relevance documentation now, because they have, like a list conversation right? And then conversation.

423 00:44:05.600 00:44:18.080 Miguel de Veyra: So I reckon what we could do is we run like a back end. Basically, we list. We done a job list conversations. And then we get those conversations. I’ve also did that before.

424 00:44:18.380 00:44:23.049 Miguel de Veyra: and then basically then analyze the entire conversation and then restore it somewhere.

425 00:44:25.000 00:44:31.379 Uttam Kumaran: Yeah, but it’s less about like, I don’t know. I just feel like that. We’re building something custom that already exists like.

426 00:44:31.620 00:44:36.359 Uttam Kumaran: I don’t want to build our own evaluator, you know.

427 00:44:37.480 00:44:46.310 Miguel de Veyra: Yeah, but I don’t think that I’ll look into it. But I don’t think there’s like one that integrates fully into, you know, relevance yet

428 00:44:46.720 00:44:49.690 Miguel de Veyra: near, even moved like their call. Next week.

429 00:44:49.970 00:44:51.459 Miguel de Veyra: Our call next week.

430 00:44:53.870 00:44:58.139 Uttam Kumaran: Yeah, I should ask. I’ll ask in the relevance channel, like, what people are doing.

431 00:44:58.410 00:45:01.210 Uttam Kumaran: But let’s take this as something to think about for next week.

432 00:45:01.660 00:45:02.210 Miguel de Veyra: Yep.

433 00:45:07.230 00:45:12.900 Uttam Kumaran: Okay, cool and then also for tool documentation.

434 00:45:15.182 00:45:18.350 Miguel de Veyra: I guess it’s I mean.

435 00:45:18.680 00:45:19.170 Uttam Kumaran: Hi, Mr.

436 00:45:19.170 00:45:24.960 Uttam Kumaran: Similar, but like we have n 8 n tools also have like relevance tools.

437 00:45:26.540 00:45:29.960 Miguel de Veyra: But honestly right now all all tools

438 00:45:30.200 00:45:35.149 Miguel de Veyra: are built under agents, right? So we don’t really like build an independent tool that won’t be used anywhere.

439 00:45:36.430 00:45:40.169 Uttam Kumaran: Oh, is that how you guys are architecting? You don’t have any tools that you’re using across agents.

440 00:45:40.450 00:45:44.080 Miguel de Veyra: No, no, but I’d rather have an agent, because it has to be conversational.

441 00:45:44.310 00:45:47.759 Uttam Kumaran: No, no, no, but like you have, like a web scraper tool.

442 00:45:50.000 00:45:52.099 Miguel de Veyra: Yeah. An agent uses that still.

443 00:45:52.330 00:45:55.460 Uttam Kumaran: But is that web scraper tool used by multiple agents.

444 00:45:58.195 00:45:59.130 Miguel de Veyra: Cause.

445 00:45:59.600 00:46:09.609 Miguel de Veyra: It depends on the use case because some web scrapers you have to be a bit more, you know, for example, the one for sasaholic. You had to give it like specific instructions for that one.

446 00:46:09.640 00:46:12.779 Miguel de Veyra: So that tool is specifically built for that one. Yeah.

447 00:46:14.740 00:46:17.049 Uttam Kumaran: So it’s now worth like documenting tools.

448 00:46:17.783 00:46:21.339 Miguel de Veyra: Unless it’s a general one, I wouldn’t. I think the agent.

449 00:46:21.480 00:46:25.509 Miguel de Veyra: I think the agents are more important. We can for tools, so we can just add, you know.

450 00:46:25.590 00:46:30.869 Miguel de Veyra: these are the tools. It’s the link is here. This is what it does. We don’t have to expand on that. I think.

451 00:46:31.140 00:46:31.900 Uttam Kumaran: Okay.

452 00:46:36.020 00:46:39.339 Uttam Kumaran: okay, great. I think that’s

453 00:46:39.900 00:46:45.460 Uttam Kumaran: the other thing I had to ask is this on the data side? So, Ryan, where are we with the synthetic data stuff.

454 00:46:47.240 00:46:48.589 Luke Daque: Yeah, I think we have a

455 00:46:49.390 00:46:52.089 Luke Daque: call with Miguel for that.

456 00:46:52.100 00:46:55.120 Luke Daque: But oh, yeah, I’ve been. Yeah. I’ve been

457 00:46:55.810 00:47:00.300 Luke Daque: testing it out for a couple of things. But yeah, I didn’t push anything yet.

458 00:47:02.060 00:47:03.280 Uttam Kumaran: It’s not deployed.

459 00:47:03.540 00:47:10.999 Luke Daque: No it, only the manufacturing one we we made together, but the others

460 00:47:11.480 00:47:19.279 Luke Daque: other categories, or like other organization types. I haven’t deployed anything yet.

461 00:47:19.750 00:47:20.370 Uttam Kumaran: Okay.

462 00:47:27.610 00:47:31.150 Uttam Kumaran: I can deploy some today. Just so

463 00:47:31.150 00:47:33.453 Uttam Kumaran: I think, let’s wait. Let’s let’s talk about like

464 00:47:33.870 00:47:38.400 Uttam Kumaran: How I mean, talk to Miguel and show him what we’re doing. But let’s we’ll create a project plan for this

465 00:47:38.640 00:47:39.570 Uttam Kumaran: as well.

466 00:47:40.970 00:47:46.759 Luke Daque: Yeah, I think we already have a notion document for that. Michael. Right? I I think I added a couple of

467 00:47:47.498 00:47:50.362 Luke Daque: things there as well like, what kind of

468 00:47:53.090 00:47:59.479 Luke Daque: What do you call that manufacturing? Or like other organization types? I’ve added as well.

469 00:47:59.670 00:48:02.989 Luke Daque: So yeah, we can work on that.

470 00:48:03.716 00:48:05.999 Luke Daque: Notion using that ocean, Docin.

471 00:48:06.380 00:48:12.149 Uttam Kumaran: Okay, okay, guys, I think that’s all I had for today.

472 00:48:12.820 00:48:15.480 Miguel de Veyra: About the automation side of things. Ulta.

473 00:48:17.242 00:48:18.450 Uttam Kumaran: What do you mean?

474 00:48:20.153 00:48:26.680 Miguel de Veyra: Like, for example, the one for you know, the target sites cause. That’s not really AI, right?

475 00:48:27.090 00:48:33.680 Miguel de Veyra: I think I should also update everyone cause. I talked to Eddie about the

476 00:48:34.670 00:48:39.440 Miguel de Veyra: basically what happened? Why, the inconsistency inconsistencies happen.

477 00:48:39.490 00:48:49.919 Miguel de Veyra: It’s you know, the selectors. And then I proposed to him Witham about the basically the screenshot thing where it’s because that will remove the dependency and the

478 00:48:50.810 00:48:52.829 Miguel de Veyra: what do you call it on the

479 00:48:53.610 00:48:58.389 Miguel de Veyra: selectors? And he’s very interested in that, but told us, basically, you know.

480 00:48:59.210 00:49:03.290 Uttam Kumaran: I mean dude. I would much rather move to doing it on image screenshot.

481 00:49:03.440 00:49:05.880 Uttam Kumaran: Didn’t I send you that screenshot Api.

482 00:49:06.940 00:49:12.830 Miguel de Veyra: Oh, yeah, yeah, I mean, it’s kinda easy to do. So it’s, you know, just connected to open AI and stuff like that.

483 00:49:13.410 00:49:15.009 Uttam Kumaran: But how would you take the screenshot?

484 00:49:16.509 00:49:21.960 Miguel de Veyra: There’s python. I’ve done it before. There’s python libraries, basically. Hey? Just take screenshot.

485 00:49:23.450 00:49:24.430 Uttam Kumaran: Oh, really. Okay.

486 00:49:24.430 00:49:30.970 Miguel de Veyra: Yeah, I’ve done it before. But yeah, he’s interested in that. And then but basically

487 00:49:31.270 00:49:36.289 Miguel de Veyra: before, cause, I think that he told me about the budget for 2025. But we didn’t really speak.

488 00:49:36.480 00:49:40.879 Miguel de Veyra: It’s just an email chain. So but yeah, and I reckon, just

489 00:49:41.210 00:49:48.859 Miguel de Veyra: improve the accuracy and then build the agent from there. Okay, off basically a demo. So we could, you know, re-sign him? Or basically.

490 00:49:48.860 00:49:53.819 Miguel de Veyra: yeah, dude, try and try this screenshot one. I feel like this could work really nicely.

491 00:49:54.020 00:49:57.789 Miguel de Veyra: Yeah, yeah, and yeah, yeah.

492 00:49:59.590 00:50:05.880 Uttam Kumaran: But that would be amazing if we can switch from doing scraping to using this. And then what you would do if you pass this to open AI.

493 00:50:06.200 00:50:22.329 Miguel de Veyra: Yep, yep, cause I’ve I think we did the cost analysis for this before I gave you something. It’s gonna cost them like 5 to $10 a day extra just to analyze it. But that at least, you know, even if something changes, we don’t have to wait a couple of days, just to make sure it’s not accurate anymore.

494 00:50:22.790 00:50:23.999 Uttam Kumaran: Yeah. Okay. Okay.

495 00:50:28.400 00:50:34.070 Miguel de Veyra: And I think they’re interested in Kroger and Walmart, too.

496 00:50:34.482 00:50:43.809 Miguel de Veyra: But they wanna get this up and running first.st But I guess the good news is, is there like Vitaco actually gave, you know, a 2025 budget for a brain forge. So.

497 00:50:44.690 00:50:46.739 Uttam Kumaran: Yeah, we’re gonna have to figure out. Yeah.

498 00:50:47.190 00:50:49.659 Uttam Kumaran: we’re figuring all that out. So okay.

499 00:50:50.350 00:50:52.180 Miguel de Veyra: And then, yeah.

500 00:50:56.200 00:51:00.839 Uttam Kumaran: Okay, cool. I think I have a lot to kind of figure out here. Basically, I wanna make sure that all of these end up

501 00:51:01.140 00:51:05.720 Uttam Kumaran: in tasks, and then we can track all of our stuff. I think we’ll I’m gonna find a way to work with

502 00:51:06.370 00:51:11.059 Uttam Kumaran: basically one to get make sure all the projects are tracked, and then I’ll work with Nico to see like, what’s the best

503 00:51:11.300 00:51:15.569 Uttam Kumaran: way to manage all this. But I’m gonna talk to him today and and think about this. So.

504 00:51:16.440 00:51:19.480 Miguel de Veyra: Do you guys wanna look into Asana, Utah.

505 00:51:20.670 00:51:22.349 Miguel de Veyra: It’s a bit expensive to be honest.

506 00:51:23.120 00:51:28.519 Uttam Kumaran: It’s not really about price for me. It’s more about another tool to manage.

507 00:51:31.290 00:51:32.230 Miguel de Veyra: I see you.

508 00:51:33.180 00:51:35.510 Uttam Kumaran: So that’s kind of like, what I’m

509 00:51:36.380 00:51:40.540 Uttam Kumaran: most worried about is like, do we want to have like multiple tools?

510 00:51:41.798 00:51:43.890 Uttam Kumaran: But maybe we could use like

511 00:51:44.520 00:51:50.309 Uttam Kumaran: I don’t know, like maybe notion. And Asana works, or or I would honestly probably try to use linear instead.

512 00:51:54.460 00:51:56.120 Miguel de Veyra: So let me think about it.

513 00:51:56.120 00:52:07.679 Miguel de Veyra: Yeah, cause I like where in? Because, remember, I cause. Remember before, we had like, we managed like 75 clients at the same time. So having, like Asana there, basically where you can see the, you know the status of this project, and then

514 00:52:07.960 00:52:13.109 Miguel de Veyra: the tasks in there when you click. It is just very helpful, I think, for notion. We could

515 00:52:13.170 00:52:19.729 Miguel de Veyra: use it a bit more for documentation than project management. I don’t know. But that’s just from what I my point of view.

516 00:52:19.930 00:52:24.019 Uttam Kumaran: Yeah, let me I think we’re gonna have to make a decision on it this month.

517 00:52:25.510 00:52:29.640 Uttam Kumaran: the thing is like, I want all of our meetings to be run out of one place.

518 00:52:30.330 00:52:33.779 Uttam Kumaran: So I’m kind of like hesitant to move out of notion.

519 00:52:37.820 00:52:43.260 Uttam Kumaran: But let’s see, I may be wrong about that, like I may may say that

520 00:52:43.420 00:52:46.269 Uttam Kumaran: we do. We can use linear or something.

521 00:52:49.930 00:52:50.510 Miguel de Veyra: I don’t know.

522 00:52:50.510 00:52:51.260 Miguel de Veyra: We’ll see.

523 00:52:51.530 00:52:52.210 Uttam Kumaran: We’ll see.

524 00:52:53.780 00:52:57.039 Luke Daque: I’ve also used Clickup before. It’s also pretty

525 00:52:57.800 00:52:59.789 Luke Daque: pretty great. And you can like

526 00:53:00.770 00:53:03.549 Luke Daque: it has like api integration to

527 00:53:03.740 00:53:07.080 Luke Daque: data warehouses as well. So we can like create our own

528 00:53:09.380 00:53:12.349 Luke Daque: visual boards or whatever data models out of them.

529 00:53:15.580 00:53:20.679 Luke Daque: But yeah, I I believe Asana has that as well, or like any other ticketing tool.

530 00:53:21.660 00:53:22.290 Uttam Kumaran: Okay.

531 00:53:28.490 00:53:32.169 Uttam Kumaran: okay, guys, anything else we want to chat about.

532 00:53:34.080 00:53:36.460 Miguel de Veyra: I think from my end, that’s pretty much it.

533 00:53:41.870 00:53:43.529 Miguel de Veyra: Yeah, I think that’s all

534 00:53:44.940 00:53:47.509 Miguel de Veyra: or data. I mean engineering stuff.

535 00:53:48.340 00:53:51.659 Uttam Kumaran: Okay, okay, cool. We’ll have another chat

536 00:53:52.840 00:53:55.059 Uttam Kumaran: next week on this stuff. And then

537 00:53:55.742 00:54:02.439 Uttam Kumaran: yeah, I think I’ll probably also, I don’t know yet whether I’m gonna organize specific AI and data meetings?

538 00:54:04.080 00:54:09.610 Uttam Kumaran: but I may end up doing that just so we can talk about specific tickets in flight as well that people need help on.

539 00:54:09.620 00:54:13.189 Uttam Kumaran: But for me we have a lot to do to make sure this is all structured. Well, so

540 00:54:13.390 00:54:15.209 Uttam Kumaran: give me some time to think about it.

541 00:54:16.230 00:54:17.000 Luke Daque: Cool.

542 00:54:17.750 00:54:18.500 Uttam Kumaran: Okay.

543 00:54:19.720 00:54:21.900 Uttam Kumaran: Alright. Guys, thank you.

544 00:54:21.900 00:54:24.330 Luke Daque: Good. Yeah, this was. This was a great meeting.

545 00:54:24.370 00:54:25.429 Uttam Kumaran: Yeah, I appreciate it.

546 00:54:25.430 00:54:30.199 Uttam Kumaran: Yeah, yeah, I’m excited for the next one already. But I have a lot to finish before then. So.

547 00:54:30.885 00:54:31.320 Miguel de Veyra: Yeah.

548 00:54:32.180 00:54:32.710 Uttam Kumaran: Okay.

549 00:54:32.710 00:54:33.320 Luke Daque: I see.

550 00:54:33.320 00:54:34.970 Miguel de Veyra: Thanks, guys, I really appreciate it.

551 00:54:34.970 00:54:35.959 Miguel de Veyra: Everyone have a good day.