Meeting Title: Uttam_Kenneth Date: 2025-03-18 Meeting participants: Ken, Uttam Kumaran


WEBVTT

1 00:00:17.120 00:00:18.500 Uttam Kumaran: Hey? Can you hear me?

2 00:00:26.600 00:00:28.660 Ken: Hello. Hello. Oh, okay.

3 00:00:29.020 00:00:30.080 Uttam Kumaran: Hey! How are you?

4 00:00:30.690 00:00:34.659 Ken: Hello! It didn’t let me adjust until after the meeting started.

5 00:00:35.000 00:00:36.330 Uttam Kumaran: No, no problem.

6 00:00:36.660 00:00:37.630 Uttam Kumaran: How’s it going

7 00:00:38.010 00:00:43.860 Ken: Hello, okay, is it working? Alright? Yeah. There we go.

8 00:00:44.450 00:00:45.310 Uttam Kumaran: Bye.

9 00:00:47.010 00:00:51.630 Ken: Wait! Volume alright, sure! There we go!

10 00:00:51.630 00:00:52.330 Uttam Kumaran: Have it now

11 00:00:52.920 00:00:53.610 Ken: Yeah.

12 00:00:54.160 00:00:56.050 Uttam Kumaran: How’s everything? Thanks for taking the time

13 00:00:56.650 00:00:57.527 Ken: Yeah, you too.

14 00:00:59.166 00:01:02.850 Uttam Kumaran: How did you? How did you and Robert get put in touch

15 00:01:04.619 00:01:06.089 Ken: He knew like,

16 00:01:07.079 00:01:12.049 Ken: okay, how old was it? He knew someone. I think he knew someone named Judy, which is basically

17 00:01:12.219 00:01:15.409 Ken: a mom of a friend I had in elementary school

18 00:01:15.410 00:01:16.589 Uttam Kumaran: Oh, really. Okay.

19 00:01:16.820 00:01:18.519 Ken: Very 6 degree, sort of thing.

20 00:01:18.854 00:01:20.859 Uttam Kumaran: Are you in la, as well

21 00:01:21.531 00:01:22.940 Ken: No, I’m in New York.

22 00:01:23.090 00:01:24.419 Uttam Kumaran: Oh, nice. Okay. Okay.

23 00:01:24.600 00:01:26.969 Uttam Kumaran: I used to live in New York. I live in Austin right now.

24 00:01:28.680 00:01:34.860 Uttam Kumaran: So, yeah, I mean, I I just wanted to chat and sort of you know, Robert mentioned that potentially

25 00:01:35.870 00:01:45.790 Uttam Kumaran: you, you may be interested in new opportunities, but would love to just share a little bit about what we’re doing. You mentioned a little bit of your background and and go and c plus plus but

26 00:01:46.231 00:02:03.989 Uttam Kumaran: I’ll just tell you a little bit about the company, and maybe we can sort of talk about some of the stuff we’re doing. So I’m the CEO of Brainforge. Brainforge is a data analytics and AI consultancy. We develop engineering solutions around data warehousing data modeling bi as well as we’re building AI agents right now.

27 00:02:04.350 00:02:04.800 Ken: Yeah.

28 00:02:04.800 00:02:09.490 Uttam Kumaran: We’ve had, you know, over like 2030 customers in our lifetime. We started the business.

29 00:02:09.639 00:02:14.680 Uttam Kumaran: and 2023 and right now we’re about like 20 people

30 00:02:15.950 00:02:27.160 Uttam Kumaran: And so growing really quickly and starting to do a lot we do a lot of stuff on the data side. A lot of lot of data platform and data engineering work. But we’re also starting to do a lot of work on AI side.

31 00:02:28.330 00:02:41.579 Uttam Kumaran: Mainly, it’s just building AI agents, doing some data, pipelining work to build context. But then also building rag systems and some integration work. So that’s really where we have a core need right now of

32 00:02:42.028 00:03:00.820 Uttam Kumaran: some help. There we have. We have 2 and a half people working on our AI team right now. But they’re both really their core competency is on like building up agents and doing a lot of the prompting side. We’re just looking for more support of people who are like interested in building AI agents and

33 00:03:00.820 00:03:01.340 Ken: Yeah.

34 00:03:01.340 00:03:03.520 Uttam Kumaran: Kind of a brand new field. So it’s

35 00:03:03.795 00:03:04.070 Ken: Correct

36 00:03:04.340 00:03:05.110 Uttam Kumaran: But yeah.

37 00:03:05.110 00:03:06.260 Ken: No, it’s really, easily.

38 00:03:06.410 00:03:09.439 Uttam Kumaran: It’s not like so complicated until like.

39 00:03:09.560 00:03:17.709 Uttam Kumaran: you want to start doing more things like fine tuning or training. But we’re at the moment we’re just building agents, and we’re doing some light pipelining, some light Api work

40 00:03:19.123 00:03:26.990 Uttam Kumaran: We’re developing solutions for ourselves internally as well, like we’re agents for ourselves. And we’re also doing stuff for clients

41 00:03:26.990 00:03:27.690 Ken: Yeah.

42 00:03:28.400 00:03:33.050 Ken: So by building agents, you mean, like, just inferencing for now, just like prompt engineering stuff

43 00:03:33.190 00:03:49.159 Uttam Kumaran: Yeah, for the most part. So exactly. So we’re doing like a lot of prompt engineering work. We’re building eval systems. We have some data work about, like about how the system’s running like more logs. But really, the the fundamental thing is, we’re trying to build a lot of integrations with slack

44 00:03:49.480 00:03:56.639 Uttam Kumaran: for our own use cases and then for clients, we’re trying to work on better rag systems like, how do you actually run retrieval over like

45 00:03:56.870 00:03:59.444 Uttam Kumaran: tons of tons of documents?

46 00:04:00.150 00:04:06.919 Uttam Kumaran: and like, how do we build? How do we build more like better chunking strategies depending on the types of data?

47 00:04:07.530 00:04:18.780 Uttam Kumaran: All that sort of stuff. I think we’ll we’ll start to do more stuff that’s around pre training around around fine tuning, and inference. But like we’re doing pretty basic stuff. Right? Now.

48 00:04:19.360 00:04:20.060 Ken: Yeah.

49 00:04:22.084 00:04:25.399 Uttam Kumaran: So yeah, I mean, we’d love to hear about like sort of what you’re working on. And

50 00:04:26.070 00:04:27.619 Uttam Kumaran: that sounds interesting.

51 00:04:28.440 00:04:29.430 Ken: Yeah, that’s pretty cool.

52 00:04:30.377 00:04:38.859 Ken: like, I’m just getting looking into, not looking to. Yeah, I’m pretty curious about AI, too. It’s like, you know, it’s like, I said. No one knows. No one really knows what we’re

53 00:04:38.860 00:04:44.790 Uttam Kumaran: No, I mean, I learned it. I just learned it myself. And it’s fun because I have a company. So I’m like, I want to automate a lot of stuff

54 00:04:44.790 00:04:45.520 Ken: Right here.

55 00:04:45.520 00:04:50.922 Uttam Kumaran: And so I. We brought on an AI team primarily just to work on our business.

56 00:04:51.640 00:05:08.300 Uttam Kumaran: But now we’re starting to actually sell some of that work to clients. And yeah, I just want us to become like, extremely competent at building agent systems. And then start to go do more things that like at the harder level, like doing more fine tuning, and then maybe even training models eventually

57 00:05:09.010 00:05:20.369 Uttam Kumaran: Which is, which is really complicated. But, like again, I want, I want us to learn a little bit more about how to do better inference strategies, better rack strategies. How to build better evals. Things like that. That’s where we’re at. Yeah.

58 00:05:20.860 00:05:24.270 Ken: Yeah.

59 00:05:24.930 00:05:28.430 Uttam Kumaran: But yeah, tell me, like what you’re up to. And like, yeah, I’m curious

60 00:05:29.860 00:05:33.159 Ken: Well, I mean, cause I like a programming. God, how do I say it?

61 00:05:33.600 00:05:38.350 Ken: And also just like, play around with programming languages like that’s not the best way to say it.

62 00:05:38.910 00:05:46.929 Ken: because I guess less about you know. For me it’s less about like a specific role that I like, but more like. I like this certain few languages that I just try to apply them to everything

63 00:05:46.930 00:05:47.340 Uttam Kumaran: Yeah.

64 00:05:47.340 00:05:50.730 Ken: You know. Review, python, go, python, and go on. My! You know.

65 00:05:50.840 00:05:56.279 Ken: How do I say it like python and go? Are the 2 that I kind of, you know.

66 00:05:56.760 00:06:11.020 Ken: just like I like to actively investigate regularly. I still have, like, you know, c plus plus and javascript, which I’m pretty strong about how do I say it? I don’t like actively, look into, you know, feel like when you have a new tool, and you want to apply to everything

67 00:06:11.020 00:06:12.150 Uttam Kumaran: Yeah, yeah, yeah, yeah.

68 00:06:13.210 00:06:17.739 Uttam Kumaran: so what? What sort of stuff are you building now? And like, like, tell me what you’re currently up to

69 00:06:18.060 00:06:26.760 Ken: Well, currently, I’m just into, you know, things like command line stuff. I don’t know if you’ve been. You’ve heard about the recent, you know, terminal, user, interface, Renaissance

70 00:06:27.610 00:06:28.550 Uttam Kumaran: No, not really

71 00:06:29.040 00:06:30.010 Ken: Yeah.

72 00:06:30.610 00:06:35.069 Ken: or it’s basically like, you know, this, every as we all know, the standard didn’t put a standard error

73 00:06:35.360 00:06:45.300 Ken: standard output. But as it turns out that you know we all the Ascii encoding. But it turns out that Ascii is actually like A, you can call it like a programming language.

74 00:06:45.300 00:06:45.940 Uttam Kumaran: Hmm.

75 00:06:45.940 00:06:57.289 Ken: Where most of it is just like, okay, print this character and then move the curse of the next character. We just use it as a data storage, but it actually everything under 30, you know, space 32, which is all the

76 00:06:57.840 00:07:01.409 Ken: we always hear about is actually meant for moving a cursor

77 00:07:01.410 00:07:05.680 Uttam Kumaran: Yeah, so, but like, are you doing that for fun? Are you doing that on a job like, what’s where

78 00:07:05.680 00:07:09.309 Ken: Just for just for fun, just for fun. Terminals. Cool. Yeah.

79 00:07:09.530 00:07:10.500 Uttam Kumaran: Nice

80 00:07:10.500 00:07:16.409 Ken: Yeah. So you know, I mean, we. There’s also like the one we know obviously, is the due line, which is, you know, move across the new line and then forward.

81 00:07:16.900 00:07:28.559 Ken: You know the infamous windows versus Linux unix line, ending, I think, the windows just ideas. You move the cursor down for the backslash end, and then backslash ours all the way forward. It’s all the way back to front.

82 00:07:29.130 00:07:38.129 Ken: So those things come from there where it’s all about, moving the cursor, and then moving the cursor home to the terminal to end, or clearing the curse, clearing the whole terminal, and that

83 00:07:38.530 00:07:38.950 Uttam Kumaran: Nice

84 00:07:38.950 00:07:46.750 Ken: If you ever heard of the things like curses, or you know and curse well, I mean, I use I’m using. Go not CC library.

85 00:07:47.030 00:07:47.620 Uttam Kumaran: Yeah.

86 00:07:47.950 00:07:51.150 Ken: Oh, God! On the top of my head! What are the

87 00:07:51.310 00:07:57.360 Ken: terminal user information we might have been using? I guess, less, you know, more or less. The the page or less

88 00:07:57.740 00:08:00.280 Ken: like that’s a terminal user interface which you know around

89 00:08:00.440 00:08:03.249 Ken: printing, paging key points, you know. Move around

90 00:08:03.440 00:08:04.080 Uttam Kumaran: Yeah.

91 00:08:04.440 00:08:13.989 Uttam Kumaran: So tell me, like, what you’re interested in doing like, does any of the AI stuff sound interesting like we’re looking for more folks that want to sort of build AI systems

92 00:08:14.180 00:08:14.760 Ken: Yeah, okay.

93 00:08:14.760 00:08:20.959 Uttam Kumaran: And for clients. So yeah, let me know if that’s any. If you have, you have you poked around at anything in that world

94 00:08:21.377 00:08:24.230 Ken: Yeah, I’ve been looking at like, what’s it called Olama? Right?

95 00:08:24.230 00:08:24.610 Uttam Kumaran: Yeah.

96 00:08:24.610 00:08:26.569 Ken: The easiest way to get started. Yeah.

97 00:08:27.120 00:08:28.949 Uttam Kumaran: And what have you done with that? So far

98 00:08:28.950 00:08:32.943 Ken: It will. I mean, just basically see how it works. Like, you know, trying like

99 00:08:34.370 00:08:43.589 Ken: normal Command line interaction normal, like the I haven’t played around with the curl with sorry not curl the cause. It has like a mode where you can. You know

100 00:08:43.909 00:08:49.899 Ken: it acts as a web server, and you could just curl into it. I’m like starting to look at. I haven’t totally

101 00:08:50.440 00:08:56.130 Ken: played around with the you know that part too much, because there is like some things like nice things you can do with, you know, saving conversations.

102 00:08:56.290 00:09:04.539 Ken: loading them in. And God, what was it called? Again the like pre pre prompting. No, there’s an actual term for it. We’re just like Pre. Write like a prompt

103 00:09:04.540 00:09:08.780 Uttam Kumaran: Oh, yeah, you have. You just have, like, a fixed prompts that are like system prompts, basically

104 00:09:08.780 00:09:11.919 Ken: Oh, yeah, that’s what it’s system prompt. So you do the system prompting.

105 00:09:12.170 00:09:15.080 Ken: There’s just a shove like a whole giant, prompt

106 00:09:15.080 00:09:15.630 Uttam Kumaran: Yeah.

107 00:09:15.630 00:09:19.109 Ken: Into it beforehand, so you can just have, you know, more accurate queries

108 00:09:19.360 00:09:20.110 Uttam Kumaran: Yeah.

109 00:09:20.110 00:09:24.600 Ken: Just shoving documents into it from the yeah, it’s pretty compatible command line actually,

110 00:09:26.630 00:09:48.719 Uttam Kumaran: So tell me, like, what your interest like was that sound interesting like, I mean, we’re we’re currently actively trying to recruit for more people doing that. I mean, I think definitely, if if I were to give some feedback would be to just take just poke around and more stuff with AI like I can definitely shoot some stuff over to you would love to see if you have any interest in like building some of those systems

111 00:09:49.100 00:09:52.360 Uttam Kumaran: like what that would look like. Yeah.

112 00:09:54.012 00:09:55.780 Ken: Yeah, that’s pretty cool

113 00:09:56.160 00:10:00.899 Ken: command line is, you know, because not just like memorizing commands. But you know the pipe

114 00:10:01.720 00:10:04.350 Ken: with the pipe operator, like, you know, piping from

115 00:10:04.730 00:10:09.029 Ken: shell, pipeline standard and standard output where it has like a pretty strong.

116 00:10:09.270 00:10:19.980 Ken: you know, pipelines. That’s it’s literally this. It’s not just like 2 different ways of saying the same. One word applied to 2 different contexts. But like, Oh, this is, they’re both in the same thing. Pipelines

117 00:10:19.980 00:10:20.620 Uttam Kumaran: Yeah.

118 00:10:20.900 00:10:24.879 Ken: So the idea with command line is, it has pretty good synergy with, you know, Llms.

119 00:10:25.060 00:10:25.470 Uttam Kumaran: Yeah.

120 00:10:25.470 00:10:28.330 Ken: That’s what I mean by the Olama, where it actually does let you

121 00:10:28.730 00:10:33.440 Ken: pipe things, you know, just dump a whole document into the query, you know, with

122 00:10:33.980 00:10:36.810 Ken: escaping. And then, yeah, it makes things easy

123 00:10:37.100 00:10:37.680 Uttam Kumaran: Yeah.

124 00:10:38.380 00:10:43.449 Uttam Kumaran: so what are you? Are you? What are you? Are you working right now? Or what’s your what is your like? What stage are you in right now

125 00:10:43.771 00:10:46.340 Ken: Right now, I’m still looking for a job.

126 00:10:47.040 00:10:48.659 Ken: Yeah, not currently employed

127 00:10:48.830 00:10:50.330 Uttam Kumaran: What did you do before this

128 00:10:52.170 00:10:52.970 Ken: Hmm!

129 00:10:53.534 00:10:58.100 Ken: Stuff like a friend’s company, where cause they were just like running a shipping company.

130 00:10:58.660 00:11:04.070 Ken: So it’s they had me, you know. Do like a couple of like automation scripts. For

131 00:11:05.350 00:11:09.862 Ken: how do you describe it? It’s basically I think it’s crud right just

132 00:11:10.210 00:11:10.920 Uttam Kumaran: Yes.

133 00:11:11.550 00:11:15.439 Ken: Dumping the data into that words. They have, like a whole bunch of places where they did regularly

134 00:11:15.860 00:11:17.900 Ken: keep track of where all the

135 00:11:18.230 00:11:26.590 Ken: what’s called the vessels are, because all the shipping companies have their own websites. They have to constantly look up where the container right now, international shipping specifically.

136 00:11:27.180 00:11:37.350 Ken: So you know all the cargo ships we have like shipping containers. They have to keep track of every single one. They’re the id numbers, but we don’t know where they are, unless you keep track of their regulars. They have to like save into a database.

137 00:11:37.940 00:11:43.679 Ken: So I have like I had like automation scripts with, you know, if you know about the auto auto hockey

138 00:11:44.360 00:11:45.279 Uttam Kumaran: Hey? Yes, yes.

139 00:11:45.280 00:11:48.319 Ken: Which is kind of like. Have you heard of playwright? I guess

140 00:11:49.470 00:11:52.550 Uttam Kumaran: Yeah. Oh, yeah, we’ve done a lot of sort of playwright and browser base.

141 00:11:52.750 00:11:57.459 Uttam Kumaran: Yeah, the browser automation. So auto hockey is like that. But it’s like for windows period. So

142 00:11:57.820 00:12:00.030 Ken: But for all windows, application.

143 00:12:00.030 00:12:03.810 Uttam Kumaran: Oh, okay, okay. So you’ve done some scripting automation work.

144 00:12:05.960 00:12:15.110 Uttam Kumaran: How about you? How about? I’ll send you a couple of things like you should you should check out this site N. 8 n dot I/O I just did in the zoom chat.

145 00:12:15.746 00:12:20.850 Uttam Kumaran: They’re like this is like what we use or I can email this over to you, too. It’s like a

146 00:12:21.858 00:12:24.909 Uttam Kumaran: it’s like an agent building framework that we use.

147 00:12:25.400 00:12:33.909 Uttam Kumaran: We use. We use Gemini and open AI for most of our stuff. And then the rest we’re using like windmill, which is like a quick python runner.

148 00:12:35.690 00:12:40.090 Uttam Kumaran: yeah, we’re just based. And then we’re using Snowflake to hold to store a lot of data snowflake and super base

149 00:12:41.840 00:12:42.410 Uttam Kumaran: Yeah.

150 00:12:44.490 00:13:02.259 Uttam Kumaran: so how about like, why don’t you take a look at some of these things and just email me, let me know if anything is interesting or where you think you might want to fit in. I mean, again, we’re we’re open to sort of bringing on people just to start for a few hours, and then, if you’re you get the hang of it, and it looks promising you like working with us. Then there’s definitely opportunity

151 00:13:02.700 00:13:05.450 Ken: Yeah, sounds, cool.

152 00:13:05.450 00:13:09.889 Uttam Kumaran: Okay, cool. Why don’t you? Yeah. How about how about like, I’ll send an email with a couple of links

153 00:13:10.395 00:13:10.700 Ken: Yeah.

154 00:13:10.700 00:13:20.130 Uttam Kumaran: And maybe just check them out and like, I don’t know. Maybe sometime later this week or next week, email me back with like your thoughts and like you played around with. And yeah, that’d be great.

155 00:13:20.740 00:13:21.580 Ken: Alright, sure!

156 00:13:21.880 00:13:25.859 Uttam Kumaran: Okay, cool. Well, you have a, you have an email. So anything else I can help with. Let me know

157 00:13:26.370 00:13:30.829 Ken: Yeah. And I’m in like New York, is that okay with you being in La or

158 00:13:30.830 00:13:34.499 Uttam Kumaran: Yeah, that’s fine. I’m actually in Austin. Robert is actually in New York.

159 00:13:34.980 00:13:35.640 Ken: Oh, okay.

160 00:13:35.970 00:13:40.540 Uttam Kumaran: Yeah, Robert’s in here. We have people in La. Here we have people all over the world, so I don’t care at all

161 00:13:40.540 00:13:42.260 Ken: So it’s a remote okay, remote. Got it

162 00:13:42.260 00:13:44.740 Uttam Kumaran: Yeah, it’s all Async remote. We’re all on slack.

163 00:13:45.553 00:13:50.109 Uttam Kumaran: So yeah, a bunch of our AI folks are actually in Asia. So

164 00:13:50.330 00:13:51.050 Ken: Yeah.

165 00:13:51.470 00:13:52.010 Uttam Kumaran: Yeah.

166 00:13:53.030 00:13:53.819 Ken: Alright, cool.

167 00:13:54.170 00:13:59.570 Uttam Kumaran: Okay, cool, alright. Well, it’s really nice meeting you. Thank you for taking the time. Yeah. And just let me know what you think

168 00:13:59.900 00:14:00.670 Ken: Sure.

169 00:14:00.910 00:14:01.789 Uttam Kumaran: Okay, appreciate it.

170 00:14:01.790 00:14:02.300 Ken: Have a nice day

171 00:14:02.300 00:14:03.449 Uttam Kumaran: Thank you. Bye.