Meeting Title: Brainforge AI Team Interview Date: 2026-03-03 Meeting participants: joaquin, Uttam Kumaran
WEBVTT
1 00:01:00.490 ⇒ 00:01:02.409 Uttam Kumaran: Hey! How are ya?
2 00:01:02.800 ⇒ 00:01:06.830 joaquin: Hey, great. Is it… how did you say your name? Utam?
3 00:01:06.830 ⇒ 00:01:07.690 Uttam Kumaran: U-Tam?
4 00:01:07.690 ⇒ 00:01:10.220 joaquin: Hey, Tom. Hey, nice to finally connect with you.
5 00:01:10.220 ⇒ 00:01:15.820 Uttam Kumaran: Yeah, nice to connect. Thank you so much for, for the patience. How’s, how was February.
6 00:01:16.100 ⇒ 00:01:29.010 joaquin: Yeah, good. I always end up kind of, like, consulting for old co-founders of mine. Yeah. They’re bootstrapping a business. They’re trying to, they work in the music industry, and they’re trying to monetize, like,
7 00:01:29.570 ⇒ 00:01:46.419 joaquin: the missed-out royalties on music, so it’s, like, $2 billion worth of royalties that are missed, every year, and so that’s kind of, like, their… their angle and their opportunity, their new, thing that they’re working on. They were the former founders of… one was a former founder of a publication place called, Elite Daily.
8 00:01:46.670 ⇒ 00:01:52.459 Uttam Kumaran: Okay. Missed… missed, payouts due to, like, people aren’t collecting, or, like, what’s the…
9 00:01:52.700 ⇒ 00:01:54.679 Uttam Kumaran: Like, wh-where- like, what is it?
10 00:01:54.870 ⇒ 00:01:58.909 joaquin: Yeah, so it’s missed royalties on plays from different platforms.
11 00:01:59.130 ⇒ 00:02:04.179 Uttam Kumaran: Oh, okay, so there’s, like, probably these edge case platforms that, like, nobody ever, like, checks in on.
12 00:02:04.180 ⇒ 00:02:12.659 joaquin: Right, it often has to do with, kind of, how you track the plays, and if things are configured on the artist side correctly.
13 00:02:12.700 ⇒ 00:02:13.880 Uttam Kumaran: Oh, okay.
14 00:02:14.140 ⇒ 00:02:27.039 joaquin: But yeah, there’s… it’s… it’s, it’s a messy place. So, I mean, it’s… infrastructure in the music industry, and some… some is… some… some records are, like, carbon copy still. So yeah…
15 00:02:27.040 ⇒ 00:02:27.640 Uttam Kumaran: I say.
16 00:02:27.640 ⇒ 00:02:28.310 joaquin: Yeah.
17 00:02:28.750 ⇒ 00:02:34.059 Uttam Kumaran: What is it like when you consult for these folks? Like, is it just architecture? Are you actually building stuff? Is it, like, both?
18 00:02:34.530 ⇒ 00:02:45.800 joaquin: So this one is mostly… so I’ve done this for them in the past, and I was kind of leading their engineering team, which they outsource. The former founder from Elite Daily, Jonathan, he’s from the Philippines.
19 00:02:45.980 ⇒ 00:03:02.079 joaquin: So we outsourced to the Philippines, so just hiring, like, the engineering team to build out. So, yeah, a little bit of hiring architecture, and then, if I have to, code. Yeah. But, you know, that’s pretty trivial now with,
20 00:03:02.080 ⇒ 00:03:07.890 joaquin: I mean, they bootstrapped with Lovable, but that can only, like, get you so far.
21 00:03:07.890 ⇒ 00:03:08.390 Uttam Kumaran: Yeah.
22 00:03:08.390 ⇒ 00:03:12.459 joaquin: So now they’re trying to, like, connect the pieces as well, so that’s kind of, like, still in progress.
23 00:03:12.630 ⇒ 00:03:14.220 Uttam Kumaran: Nice, nice. Okay.
24 00:03:14.960 ⇒ 00:03:20.650 joaquin: How about yourself? How do you… how do you know Vinnie? He’s really excited about what you’re doing, and .
25 00:03:20.650 ⇒ 00:03:29.940 Uttam Kumaran: I appreciate that. I got put in touch with him through a mutual friend of ours, whose name is Clarence. He worked at EY as well, and they worked together.
26 00:03:30.250 ⇒ 00:03:34.649 Uttam Kumaran: And yeah, like, I… I don’t know, I… I don’t have a ba…
27 00:03:34.650 ⇒ 00:03:39.880 joaquin: background in consulting. I’ve worked as a data engineer, I led data teams for a while.
28 00:03:39.880 ⇒ 00:03:42.580 Uttam Kumaran: A lead product for a while,
29 00:03:43.030 ⇒ 00:03:57.059 Uttam Kumaran: And when I got into this world, sort of just, like, make friends along the way. Like, I learn a lot about consulting just from speaking to people that are at various levels, and, got in touch with him and sort of shared with him a little bit what we’re doing on the AI side, hearing about
30 00:03:57.070 ⇒ 00:04:04.310 Uttam Kumaran: what he’s trying to do, and then usually closing, I always mention to people, like, hey, if not…
31 00:04:04.350 ⇒ 00:04:15.300 Uttam Kumaran: you know, the person I’m talking to, I say, if not yourself, like, if you have anybody in your life that, you know, is interested in making a switch, or working with cool people on, sort of, cutting-edge stuff, I would love to
32 00:04:15.330 ⇒ 00:04:24.390 Uttam Kumaran: you know, chat with them, and so usually that gets a couple bites, but yeah, so that’s sort of how we got connected. I tell them a lot about,
33 00:04:24.530 ⇒ 00:04:29.539 Uttam Kumaran: You know, for folks that are in big consulting, really what I talk to them about is how we’re trying to run our…
34 00:04:29.700 ⇒ 00:04:41.369 Uttam Kumaran: consulting company, like, really, really AI native. Of course, we also are selling AI services, like AI, related application, like workflow automation, but we…
35 00:04:41.440 ⇒ 00:04:50.680 Uttam Kumaran: run our company super, super AI-forward. So, usually that’s what they’re interested about in the consulting world, like, how are you getting teams to adopt, or how are you using it
36 00:04:51.020 ⇒ 00:05:08.819 Uttam Kumaran: beyond just, like, engineers are using Cursor. And, like, really, like, that’s also, like, kind of the M.O. of our company, like, I started the business really focused on data engineering, data modeling, like, kind of, like, building basically a data agency, and then we used AI a lot, and I actually built…
37 00:05:08.820 ⇒ 00:05:20.960 Uttam Kumaran: I actually started hiring AI, or, like, kind of full-stack folks who were, like, interested in AI to help me build more of our internal platform. And then naturally, we were, like, a year or two ahead of the curve on a lot of stuff.
38 00:05:20.980 ⇒ 00:05:31.699 Uttam Kumaran: So then now we’re actually able to take a lot of what we learned and, like, now go sell AI services, so transformation-related services, which has been a blast. And still, I think, internally, we continue
39 00:05:31.850 ⇒ 00:05:49.730 Uttam Kumaran: in many ways, not only, I think, to be on the edge, in terms of, like, probably what is happening in technology, but in consulting, it’s so behind. So we’re oftentimes, like, we… I think more of our service is like a product, and internally, we run really a lot closer to product team than we do run.
40 00:05:49.730 ⇒ 00:05:52.899 joaquin: You know, than a consultancy. Yeah.
41 00:05:52.900 ⇒ 00:05:54.590 Uttam Kumaran: That’s… that’s cool. Yeah.
42 00:05:54.970 ⇒ 00:06:10.450 joaquin: what is, like, the, the bread and butter service that you… that you offer? It’s like, okay, like, these… these are what, like, the companies are really interested in, like, this is kind of, like, the solution you provide them. Or is it just, like, an analysis of their process, and, like, you provide, like,
43 00:06:10.470 ⇒ 00:06:18.389 joaquin: an array of different possibilities for them, right? Or are you kind of… Integrating, into their platform.
44 00:06:18.540 ⇒ 00:06:26.349 Uttam Kumaran: Yeah, so, like, our company split into, like, 3 core services, so we have, like, data, we have strategy and analytics, and then we have AI.
45 00:06:26.360 ⇒ 00:06:27.270 joaquin: So…
46 00:06:27.270 ⇒ 00:06:37.190 Uttam Kumaran: our, you know, our company was built all on, you know, data strategy and analytics. So this is setting up data warehouses, landing data, data modeling.
47 00:06:37.270 ⇒ 00:06:56.620 Uttam Kumaran: BI tool implementation, and then also all the way up to, like, insights and strategy, like, you can think about, like, PE-style decks and things like that. So, that’s what I’ve done my whole career, and, like, that’s what our… most of our team does. On the AI side, it really came out of, like, one, which is, like, everything related to internal workflow automation.
48 00:06:56.730 ⇒ 00:06:58.340 Uttam Kumaran: Now we’re also continuing.
49 00:06:58.650 ⇒ 00:06:59.449 joaquin: Or is this…
50 00:06:59.450 ⇒ 00:07:11.110 Uttam Kumaran: We started doing some NA to N work maybe, like, 2 years ago. Now it’s all, like, custom builds. So we’re using, like, Mastra, which is, like, different… so you could use a multitude of various frameworks for
51 00:07:11.110 ⇒ 00:07:27.290 Uttam Kumaran: for AI agents, most of it is actually a lot of just, like, a lot of data engineering, like, moving contacts to the right place, and then building whatever, UX or UI is necessary for the automation. We also do a lot of, like, Slack work, and work into, like, various different integrations.
52 00:07:27.290 ⇒ 00:07:30.839 joaquin: But again, it’s… nothing is customer-facing, so it’s all internal.
53 00:07:30.840 ⇒ 00:07:44.109 Uttam Kumaran: So it’s all workflow automation, it’s all, like, diagnosing a problem, doing a time study, understanding, like, how can we build some workflows to do that, whether that’s within tools, whether that’s something externally, but also a lot of the work that we’re starting to do as really
54 00:07:44.110 ⇒ 00:07:54.230 Uttam Kumaran: more of, like, how do you actually do… how do you actually create, like, a knowledge layer across an enterprise? Like, in our company, we have tons of different hookups into, like.
55 00:07:54.230 ⇒ 00:08:09.719 Uttam Kumaran: meeting transcriptions, HubSpot, Slack, and so at our company, you can… someone can go into Cursor and ask a question, and it’ll pull from a variety of resources to help them. Oh, wow. And that’s something that a lot of companies are like, holy shit, we need that. And…
56 00:08:09.720 ⇒ 00:08:16.089 Uttam Kumaran: I think we’re in a unique spot, because we’re, like, a data company, so, like, these are all data engineering problems, like, I…
57 00:08:16.100 ⇒ 00:08:24.509 Uttam Kumaran: I didn’t know that this is actually what a lot of the industry is struggling with now, because the LLMs have all caught up to where I thought they would be, like…
58 00:08:24.550 ⇒ 00:08:28.699 joaquin: And those will keep getting better, but without the proper context at the right time.
59 00:08:28.700 ⇒ 00:08:43.159 Uttam Kumaran: It’s… doesn’t matter whether you’re using GPT-4 or whether you’re using Opus, it’s like, you need to have the right context available. And then also, like, again, even… I think the… for the average, like, knowledge worker.
60 00:08:43.320 ⇒ 00:08:52.119 Uttam Kumaran: like, in engineering, we have… we have Cloud Code and Cursor, but most folks in the business don’t have a UI for the stuff that they do, so…
61 00:08:52.280 ⇒ 00:08:55.339 Uttam Kumaran: I feel like in my company, I’ve had the luxury and…
62 00:08:55.390 ⇒ 00:09:11.770 Uttam Kumaran: I think, of course, because I’m sort of leading, I’m able to… basically, everybody uses Cursor, including our salespeople, marketing people, everybody, no matter what. And so I… but I know that it’s difficult for them, and so we’ve trained, and we’ve, like, shown them that this is actually, like, yes, it’s like an IDE,
63 00:09:11.800 ⇒ 00:09:19.180 Uttam Kumaran: But actually, it’s becoming less and less than that. Right, right. But for the general public, people want UI still.
64 00:09:19.210 ⇒ 00:09:36.330 Uttam Kumaran: Right? And so we’re finding it interesting, like, building those types of experiences for folks. And then also, it’s like, even on the data side, right, we’re using AI not only to deliver all of our services, but then even within the data side, a lot of what we’re deploying for clients is, like.
65 00:09:36.380 ⇒ 00:09:55.370 Uttam Kumaran: sort of, like, like, data analyst-related agents, something that can analyze data, pull out insights, do that on some recurring basis. And so, like, it’s kind of permeated everywhere. I think our real challenge right now is, frankly, continues and always has been just, like, finding great people, and…
66 00:09:55.420 ⇒ 00:10:09.660 Uttam Kumaran: putting them in the best position to deliver for clients. And so, like, ultimately, like, our business is… is like a broker. Like, in the… if you take it to its peak, right? We find really, really amazing people, and then we pair them with, like, really, really amazing problems.
67 00:10:09.690 ⇒ 00:10:21.759 Uttam Kumaran: And I think typical consultancies and tech consultancies, they, like, kind of lost sight of, like, I think, people a lot. I’ve worked as a contractor and as a consultant. Like, freelance is really, really hard.
68 00:10:21.830 ⇒ 00:10:35.340 Uttam Kumaran: It, like, it could quickly be like, oh, I went from, like, being inside a company, now I have 3 bosses, and, like, no healthcare. So I do understand that, like, from the origination of this company, we wanted to be, like, a consulting firm.
69 00:10:35.480 ⇒ 00:10:43.019 Uttam Kumaran: But it took a while. We’re, like, completely bootstrapped, so it took us… it’s almost been 3 years now since I quit my last job.
70 00:10:43.020 ⇒ 00:10:44.849 joaquin: Yeah, you’re essentially a startup, essentially, you know?
71 00:10:44.850 ⇒ 00:10:54.579 Uttam Kumaran: Yeah, but I think… I think this is where, like, this is also where, though, like, we are… I’ve worked in many venture-backed startups, like, we are very, very different than that, like.
72 00:10:54.580 ⇒ 00:10:54.900 joaquin: It’s not…
73 00:10:54.900 ⇒ 00:11:05.899 Uttam Kumaran: not crazy, we’re not burning tons of money, there’s not… there’s not, like, a… I’m working every day to make sure that there is, like, a 90% chance we succeed. Like, this isn’t like a…
74 00:11:06.070 ⇒ 00:11:12.460 Uttam Kumaran: Oh, like… maybe this is gonna work, like, this is a tried and true model. Instead, the game…
75 00:11:12.460 ⇒ 00:11:13.540 joaquin: hard to fit, right.
76 00:11:13.540 ⇒ 00:11:29.250 Uttam Kumaran: Yeah, the games I play are with, like, can we use AI to deliver faster? Can I run it, like, actually, like, a better company for the engineers, where you’re not, like, stuck on one client, you don’t know anything, like, about the rest? Like, can I run it more like a product company style?
77 00:11:29.370 ⇒ 00:11:47.299 Uttam Kumaran: but also deliver at the product company pace, but our delivery is a service, right? But again, now you’re hearing a lot about these, like, AI-native service companies. Like, when I started, it wasn’t, like, 2 or 3 years ago, it wasn’t really that popular of a mindset. I just had constraints.
78 00:11:47.510 ⇒ 00:12:00.289 Uttam Kumaran: Like, I just didn’t have any money. So, we slowly, slowly built it up, but that’s sort of, like, where we’re at now. So, we’re about 25 people now. Oh, wow. And just, like, kind of continuing to grow, you know?
79 00:12:00.540 ⇒ 00:12:06.739 joaquin: Yeah, awesome. I really like the, the whole… because so information… there’s so much information is sparse.
80 00:12:06.840 ⇒ 00:12:14.149 joaquin: across the organization, it sounds like you’re kind of centralizing it, and leveraging a model through Cursor.
81 00:12:14.150 ⇒ 00:12:20.420 Uttam Kumaran: to have a single interface to interact with all the information across the organization. Correct. Kind of like the…
82 00:12:20.460 ⇒ 00:12:26.540 joaquin: the solution for that? Like, how do you unify all the information across your different systems?
83 00:12:26.540 ⇒ 00:12:37.579 Uttam Kumaran: Yeah, so internally at our company, I’ve kind of made the exact… I mean, we’ve been building our, like, an internal platform, which has been a mix of UI or, like, playbooks, right? If you consider, like, the Brainforge platform.
84 00:12:37.580 ⇒ 00:12:48.099 Uttam Kumaran: It’s a mix of tools for everybody to use internally, externally. It’s maybe, like, a bunch of, like, helpful scripts or code, but then also we built a UI. So we’ve had, like, an internal UI to help us manage clients.
85 00:12:48.120 ⇒ 00:12:52.810 Uttam Kumaran: Like, we record a lot of our meetings, so you can go rewatch meetings, you can chat over them.
86 00:12:52.870 ⇒ 00:13:04.829 Uttam Kumaran: And so… but the kind of the problem is we built it sort of, like, monorepo, so kind of the big thing in the last two months is, like, we just, like, shoved it all into one. We previously were in a monorepo, so we shoved it all into one.
87 00:13:05.880 ⇒ 00:13:20.320 Uttam Kumaran: And then the second thing was I… I… we’ve trained every single person in our company to use Cursor, and so now they can use something sitting on top of, like, a monorepo to ask questions about, like, how do I write an SOW?
88 00:13:20.370 ⇒ 00:13:33.050 Uttam Kumaran: When’s the last time we discussed this product? Like, who should I ask for for this question? It’s much… it’s much faster, and the quality is so much better than, like, ChatGPT hooked into, like, Google Drive, like…
89 00:13:33.050 ⇒ 00:13:36.409 joaquin: That is, like, so far from, like, where we are now.
90 00:13:36.410 ⇒ 00:13:44.060 Uttam Kumaran: So part of this is I’m also deciding, I’m like, okay, what needs to be… like, what needs to be saved in files versus what needs to be…
91 00:13:44.210 ⇒ 00:13:46.540 Uttam Kumaran: In a database where, like.
92 00:13:46.660 ⇒ 00:14:06.119 Uttam Kumaran: you can query, and… because you want to traverse across multiple… and then what needs to, like, what is still required to be an MCP, and then what do we need to use CLI for, right? So, you have these various modes of, like, accessing information. I… I feel like we’re still aligning on, like, what the…
93 00:14:06.210 ⇒ 00:14:25.819 Uttam Kumaran: plays are, depending on the type of data. Like, for example, we have… we ingest all our Slack messages, but me throwing all that into the repo, it’s like, doesn’t… I don’t think that’s… right? But I’m like, okay, so I need to create a Markdown file, though, that instructs everybody’s cursor how to access black messages in Supabase.
94 00:14:25.860 ⇒ 00:14:30.200 Uttam Kumaran: Using, like, just any sort of Postgres SQL, right?
95 00:14:30.310 ⇒ 00:14:39.550 Uttam Kumaran: And so that’s actually, like, what we’re… what we’re doing there, is, like, we store all of it in a Postgres DB. Everybody in Cursor, if you have… if you pull down our platform repo.
96 00:14:39.720 ⇒ 00:14:54.499 Uttam Kumaran: the instructions are there, so you don’t have to think, how do I write this? It’s like, everybody in the company can just say, like, I want to check these Slack messages. There’s already a markdown file with instructions on, like, how to do that. So Cursor will automatically do that. So my job here…
97 00:14:54.640 ⇒ 00:14:56.959 Uttam Kumaran: Is thinking about these, like, platforms.
98 00:14:56.970 ⇒ 00:14:59.800 joaquin: giving people… making sure people are using Cursor.
99 00:14:59.800 ⇒ 00:15:19.370 Uttam Kumaran: And then continuing to just build a foundation underneath them, so that they can eventually come to Cursor and be like, hey, my role on this team is this? Like, tell me what I need to do today, right? And from transcripts, from our typical SOWs, from our project plans, it’s really clear, you know, and they can do that work.
100 00:15:20.040 ⇒ 00:15:25.290 joaquin: Very cool. And, like, what, do you have software engineers, at your company?
101 00:15:25.530 ⇒ 00:15:30.739 Uttam Kumaran: Yeah, we have, yeah, we have a bunch. Yeah, I mean, we have… most of the company’s engineers.
102 00:15:30.740 ⇒ 00:15:32.850 joaquin: Oh, wow. There’s, like, yeah, so…
103 00:15:32.860 ⇒ 00:15:34.320 Uttam Kumaran: I feel like I’m…
104 00:15:34.490 ⇒ 00:15:38.309 Uttam Kumaran: I’m one of the only, like, few… pure… I’m one of the only, like, business people.
105 00:15:38.310 ⇒ 00:15:42.260 joaquin: Like, me, my business partner, Robert, but both of us are data people.
106 00:15:42.260 ⇒ 00:15:47.720 Uttam Kumaran: So, on our AI team, we have about 5 people, and on the data team, there’s, like.
107 00:15:48.390 ⇒ 00:15:50.089 Uttam Kumaran: 10 or 11 people?
108 00:15:50.450 ⇒ 00:15:52.999 Uttam Kumaran: And then the rest, like, we have, like, 2 people on ops.
109 00:15:53.320 ⇒ 00:15:55.799 Uttam Kumaran: One person who kind of leads recruiting.
110 00:15:56.020 ⇒ 00:16:02.470 Uttam Kumaran: And then… Like, a couple designers, and then some people in marketing.
111 00:16:02.810 ⇒ 00:16:14.740 joaquin: Okay. And the, the people in AI, are they, like, AI engineers, or are they software engineers who, are familiar with enterprise AI? Like, or are they, like, AI scientists?
112 00:16:15.030 ⇒ 00:16:29.589 Uttam Kumaran: Yeah, no, it’s all… so, I would say, like, AI engineers, of course, is just, like, kind of a made-up word, I feel like. Like, if you were really an AI engineer, I think you’d just be at Facebook, like, making a bajillion dollars. But for the most part, these are, like, full-stack folks, or back-end folks.
113 00:16:29.590 ⇒ 00:16:31.279 joaquin: who I found…
114 00:16:31.710 ⇒ 00:16:50.369 Uttam Kumaran: we’re, like, interested in AI, and I’m like, I have AI problems, but of course, like, actually, like, picking an LLM and stuff is, like, the easiest part of most of what we do. It’s actually a lot of, like, backend stuff, it’s a lot of API development, and then it’s some, like, front-end UX for, like, chat interfaces or various interfaces.
115 00:16:50.370 ⇒ 00:16:52.850 joaquin: So for the most part, they’re all, like.
116 00:16:52.850 ⇒ 00:17:00.310 Uttam Kumaran: one of those, various, like, levels, most… but our crew in Brainforge, like, it’s mostly mid-level to senior people, like, we don’t have…
117 00:17:00.450 ⇒ 00:17:03.089 Uttam Kumaran: very junior people.
118 00:17:03.290 ⇒ 00:17:08.629 Uttam Kumaran: I think even we have people that are earlier in their career, but are, like, nasty. And so…
119 00:17:09.010 ⇒ 00:17:19.689 Uttam Kumaran: we… I would say, like, as we grow, we’ll probably think more about, like, bringing on junior people and growing, but for the most part, we’re just looking for people that can, like, hit the ground running for clients.
120 00:17:19.690 ⇒ 00:17:25.599 joaquin: And that team is growing, like, we have a bunch of new AI work that’s coming, and we’re starting to do a lot of work with…
121 00:17:25.599 ⇒ 00:17:30.910 Uttam Kumaran: like, Snowflake, with some AI stuff. We use, like, all the most common…
122 00:17:31.160 ⇒ 00:17:51.059 Uttam Kumaran: sort of things in AI, so we do a lot of, like, eval work, we’re building chat interfaces, we’re building, like, we built, like, Google Chat-related agents. We have… we’re internally, you know, we’re… we’ve been testing out a lot of, like, cursors, like, cloud agents this past week. And so, like, a lot of that, really, I think where our work is gonna…
123 00:17:51.400 ⇒ 00:18:07.589 Uttam Kumaran: like, where we’re finding product-market fit in terms of our service is, like, helping companies structure their data in a way that it can be accessed via AI and used for knowledge work. Like, if I have to tie a bow on the whole thing, that’s how both parts of our world sort of intersect.
124 00:18:07.620 ⇒ 00:18:13.269 Uttam Kumaran: And then really, it’s like, we’re telling the story of, like, how that’s driving ROI for… for one of our clients.
125 00:18:13.450 ⇒ 00:18:17.009 joaquin: For sure. I do… I do think that that’s, like, a problem that needs to be solved for…
126 00:18:17.010 ⇒ 00:18:17.610 Uttam Kumaran: Yeah.
127 00:18:17.820 ⇒ 00:18:23.009 joaquin: So, that makes total sense. What’s kind of, like, the stack over there? Like, what,
128 00:18:23.230 ⇒ 00:18:25.420 joaquin: programming languages is… does your…
129 00:18:25.420 ⇒ 00:18:29.530 Uttam Kumaran: Yeah, our team right now is using a lot of TypeScript. Okay.
130 00:18:29.650 ⇒ 00:18:33.340 Uttam Kumaran: Yeah, so they’re using a lot of TypeScript, so it’s all sort of, like, JavaScript.
131 00:18:34.510 ⇒ 00:18:45.910 Uttam Kumaran: we don’t do a lot of, like, Python work on the AI side, and then, in terms of backend, it sort of depends. Like, if a client… if a client is on AWS, then we have to work there.
132 00:18:45.950 ⇒ 00:18:49.719 joaquin: I think, again, like, it’s becoming less and less relevant, because.
133 00:18:49.760 ⇒ 00:19:01.839 Uttam Kumaran: you could use the CLI, and you can kind of do whatever config, but yeah, I would say, like, I don’t know what we’re… again, it really just depends, so we… our stack internally is very different than what we’ve…
134 00:19:01.980 ⇒ 00:19:14.880 Uttam Kumaran: may or may not have done for clients, but it’s nothing… it’s all cloud-native stuff. Right. So it’s nothing too complicated. I feel like I would be really… I’m hopeful that we can start doing more complex work in terms of fine-tuning
135 00:19:14.880 ⇒ 00:19:22.479 Uttam Kumaran: and hopefully, like, training, as well as I hope we can go deeper on, like, actually running evals more seriously, and building
136 00:19:22.500 ⇒ 00:19:35.400 Uttam Kumaran: building agent… building agents that we can actually, like… we do a lot of measurement work when we ship AI, like, how often this is running, who’s using it, how many tokens, what the cost is, but I’d like us to go deeper there as well, and then I think eventually, like.
137 00:19:36.230 ⇒ 00:19:45.600 Uttam Kumaran: internally, I’m… I’m working with our team on, like, how can we leverage, like, codecs and cursor agents to speed up our development time, right? Create…
138 00:19:45.680 ⇒ 00:19:58.390 Uttam Kumaran: I’ll be able to assign out tasks to agents to execute, and have our time be more probably focused on reviewing code, or thinking about, like, broad architecture, versus, like, too much hands-on keyboard.
139 00:19:58.390 ⇒ 00:20:00.050 joaquin: Right. But again, we’re just, like.
140 00:20:00.330 ⇒ 00:20:01.529 Uttam Kumaran: we’re just…
141 00:20:01.870 ⇒ 00:20:15.719 Uttam Kumaran: pushing that, our narrative internally, and so, like, the clients we’re working with are so far behind us. So oftentimes, they’re, like, they’re just like, oh, I try… we deploy ChatGPT to everybody, but, like, there’s hiccups in, like, hooking everything up.
142 00:20:15.970 ⇒ 00:20:23.489 Uttam Kumaran: That’s, like, where we were, like, a year ago, or, like, a year and a half ago. So that’s, like, the class of problems for our clients is way, way simpler than…
143 00:20:23.730 ⇒ 00:20:28.499 Uttam Kumaran: the stuff we’re doing internally, but I think it’s nice at this company because we get to dog food a lot of it.
144 00:20:28.740 ⇒ 00:20:36.179 Uttam Kumaran: I mean, we’re going to other consultancies and being like, we just solved this for ourselves, like, we’d love to share this with you, you know, so…
145 00:20:36.290 ⇒ 00:20:41.219 Uttam Kumaran: And then the team is really awesome, like, I feel like we’re… because,
146 00:20:41.430 ⇒ 00:20:50.610 Uttam Kumaran: like, I think because me, like, I’m a… my background’s in engineering, a lot of the culture is, like, really friendly for, like, engineering work. Like, everybody’s fully remote.
147 00:20:50.810 ⇒ 00:20:52.170 Uttam Kumaran: Like…
148 00:20:52.300 ⇒ 00:21:00.159 Uttam Kumaran: we don’t have a ton of meetings. It’s really, like, really, I think the complexity comes from, like, client and client communications.
149 00:21:00.300 ⇒ 00:21:05.680 Uttam Kumaran: the work isn’t that… isn’t that crazy. I think it’s always just related to, like, how do we communicate better?
150 00:21:05.840 ⇒ 00:21:13.369 Uttam Kumaran: And then how do I… how do… how do client pods, like, execute well, and how do we set the incentives well enough for them to do so, you know?
151 00:21:13.870 ⇒ 00:21:20.079 joaquin: Is that how it’s structured? It’s, each client will have, like, essentially a dedicated team?
152 00:21:20.080 ⇒ 00:21:31.489 Uttam Kumaran: Yeah, so right, so there are a few clients that have, like, dedicated folks, but most of the people now are split between, like, 2 or 3 clients. Yeah. Sort of depending on, like, what the contract is and resourcing.
153 00:21:31.610 ⇒ 00:21:51.450 Uttam Kumaran: I think as we’re growing, we’re getting bigger and bigger clients to where I actually do hope that things will get more dedicated, because contact switching can be hard. But yeah, usually people now are on at least 2 or 3 clients, and again, like, what we’ve built in terms of a platform is all these ways for
154 00:21:51.460 ⇒ 00:21:56.899 Uttam Kumaran: For us to move from… as previously, we would have, like, client pods of, like, at least 5 people.
155 00:21:57.010 ⇒ 00:22:02.429 Uttam Kumaran: One person was an account manager, one was a project manager. We basically, like, eliminated that.
156 00:22:02.510 ⇒ 00:22:15.939 Uttam Kumaran: And now the engineers on a project run all of the project management, run all the client presentations, and it’s a minimum of 3. And so it’s really nice, and we have a little bit of a structure on how we do that, but we’ve built a lot of helpful scaffolding
157 00:22:15.980 ⇒ 00:22:24.830 Uttam Kumaran: for a pod to, like, run an engagement. Not only the development work, but all the client communications, project management.
158 00:22:24.990 ⇒ 00:22:38.009 Uttam Kumaran: And then our clients are growing, so we… meaning the average revenue of our client is growing, and I think the amount of revenue that they’re driving us is starting to grow. And so we’re gonna be able to staff a little bit bigger teams, and hopefully work with
159 00:22:38.430 ⇒ 00:22:39.190 Uttam Kumaran: like…
160 00:22:39.690 ⇒ 00:22:55.200 Uttam Kumaran: fewer, but larger people, probably longer term. But we’re kind of growing in every angle, like, we still get a lot of clients that are smaller, that we’re like, okay, we should have a team of three, but maybe it’s just, like, 20 hours a week of work. And then we still have… we have larger clients where there’s, like, 7 people on.
161 00:22:55.340 ⇒ 00:22:56.959 Uttam Kumaran: On a given client, you know?
162 00:22:57.450 ⇒ 00:23:06.349 joaquin: what’s the pipeline… what’s, like, the bottleneck in the pipeline? Is it, like, do you have clients on… on waiting right now? Like, what is kind of, like, that…
163 00:23:06.350 ⇒ 00:23:19.909 Uttam Kumaran: Yeah, I mean, right now, the bottleneck is, like, my business partner, Robert, and I are still doing client work, and so we have active work on clients that we are turning down because we don’t have enough people.
164 00:23:20.050 ⇒ 00:23:25.160 Uttam Kumaran: So, part of it is… is… is staff. So, We’ve, we’ve never,
165 00:23:25.270 ⇒ 00:23:41.610 Uttam Kumaran: we’ve tried very, very hard not to lower the bar when demand comes in. Like, the worst thing I can do is just, like, be like, cool, like, we need people, let’s just find people and throw… like, we just don’t do that, so we’ll turn down work until we can satisfy it, but we have a lot of demand for our work.
166 00:23:41.690 ⇒ 00:23:46.829 Uttam Kumaran: Right now, the main bottleneck is moving me and Robert out of…
167 00:23:46.920 ⇒ 00:23:51.409 Uttam Kumaran: Day-to-day client work, so we can go sell The next biggest contract?
168 00:23:51.440 ⇒ 00:23:53.399 joaquin: And so that I can go sell…
169 00:23:53.400 ⇒ 00:23:55.770 Uttam Kumaran: like, these, like, frontier AI solutions.
170 00:23:55.770 ⇒ 00:23:56.530 joaquin: that, like.
171 00:23:56.590 ⇒ 00:23:59.809 Uttam Kumaran: to enterprise, which takes a long, long time.
172 00:24:00.080 ⇒ 00:24:17.820 Uttam Kumaran: But, like, we have such a clear path to do it. Like, we’re… we’ve gone from, like, zero to, like, where we are now with absolutely, like, nothing, and we’ve built all of our marketing, all of our go-to-market, all of our sales, and we have a clear path into that, and we’re growing really, really fast.
173 00:24:18.040 ⇒ 00:24:25.610 Uttam Kumaran: And so my job is just to basically remove myself, right? And it’s partly painful because I, like, love.
174 00:24:25.610 ⇒ 00:24:26.760 joaquin: of this work.
175 00:24:26.760 ⇒ 00:24:44.750 Uttam Kumaran: But at the same time, I still do a lot of, like, I’m… I have Cursor up in, like, four windows, still doing a lot of stuff, but it’s my job to, like, shine the flashlight into, like, the area where we have… the company hasn’t gone, and, like, be the first person, or… and go sell the biggest deal in, like, the most, like.
176 00:24:44.980 ⇒ 00:25:02.209 Uttam Kumaran: tense environment, and then make sure everybody can make as much money as possible around here. So that’s what the company, like, needs me to do, and we actually, like, have a list of, like, all the responsibilities in the company, and basically, like, for our leadership crew, it’s like, you know, we just have to eliminate my name from as many of those.
177 00:25:02.330 ⇒ 00:25:06.569 Uttam Kumaran: So that it can free me up to go, like, okay, if we want to go after Fortune 500, like.
178 00:25:06.760 ⇒ 00:25:20.670 Uttam Kumaran: okay, it’s gonna take, like, a lot of work to get there. And so, that’s the next, like, largest thing that we’re working on, is really, like, moving us out of delivery. But, like, to tell you the truth, like, it’s working, like, even today.
179 00:25:20.840 ⇒ 00:25:30.910 Uttam Kumaran: It hasn’t closed yet, but we may close our first deal that doesn’t involve either of us. Like, someone who’s new on our sales and marketing team went with another one of our senior engineers.
180 00:25:30.910 ⇒ 00:25:31.600 joaquin: Oh, wow.
181 00:25:31.600 ⇒ 00:25:39.399 Uttam Kumaran: found, like, a… it was one of his past contacts, and they sold the deal directly without including us at all. And so that was, like, a watershed moment, like…
182 00:25:39.520 ⇒ 00:25:41.570 Uttam Kumaran: Where I’m like, okay, we’ve built…
183 00:25:41.970 ⇒ 00:25:46.309 Uttam Kumaran: We’ve built the structure in a way that, like, it could start to run.
184 00:25:46.610 ⇒ 00:26:03.900 Uttam Kumaran: But again, like, we’ve all… I’ve thought about this company as one big engineering problem. Like, we have to engineer these, like… for someone like Luke on my marketing team to, like, sell that, okay, he needs to know how to, like, write SOWs, he needs to know what all our talking points are. So, like, all of that needs to be put into AI context.
185 00:26:03.950 ⇒ 00:26:06.379 joaquin: And so he can actually learn a lot faster.
186 00:26:06.380 ⇒ 00:26:10.769 Uttam Kumaran: He’s actually shipped his own Vercel demo for this demo that he did today.
187 00:26:10.890 ⇒ 00:26:16.579 Uttam Kumaran: And that’s the stuff I’m like, okay, cool, like, if we’re doing that, I see the light at the end of the tunnel.
188 00:26:16.750 ⇒ 00:26:31.959 Uttam Kumaran: You know, and then really my time’s just gotta be spent, one, on, like, recruiting the best people. It’s recruiting the best people, retaining those people, and then going after the next biggest deal, right? That’s, like, where the com- as an employee.
189 00:26:32.100 ⇒ 00:26:34.670 Uttam Kumaran: That’s where the company needs me to… to go.
190 00:26:34.940 ⇒ 00:26:39.949 joaquin: Right, shifting more towards, like, a CEO role than someone in the weeds.
191 00:26:39.950 ⇒ 00:26:40.550 Uttam Kumaran: Yeah.
192 00:26:41.600 ⇒ 00:26:48.600 joaquin: Cool, so how is it that I can understand better how we… if there’s, like, an opportunity for us to… to work together?
193 00:26:48.600 ⇒ 00:26:57.470 Uttam Kumaran: Yeah, I mean, tell me what you’re thinking, like, what are you thinking about in your career? Like, what’s next? Like, what’s interesting about this opportunity, like, from some of the things that you’ve been hearing?
194 00:26:57.940 ⇒ 00:27:02.530 joaquin: Yeah, so I mean, one of my big, biggest focuses that I haven’t
195 00:27:02.640 ⇒ 00:27:19.509 joaquin: got a chance to spend much energy on is AI, but not just, like, the surface-level AI of NAN or, like, you know, ChatGPT, Cursor, things like that. I’m trying to go, even a little bit deeper. There’s this course called,
196 00:27:19.830 ⇒ 00:27:21.600 joaquin: Deep Atlas, let me share this.
197 00:27:21.600 ⇒ 00:27:22.210 Uttam Kumaran: Yeah.
198 00:27:22.400 ⇒ 00:27:25.340 joaquin: this course called Deep Atlas, that…
199 00:27:25.790 ⇒ 00:27:28.879 joaquin: I am, starting the program, too.
200 00:27:29.130 ⇒ 00:27:38.760 joaquin: And they essentially teach you how to build LLMs from scratch, deploy these into production, build agentic AI from scratch.
201 00:27:38.880 ⇒ 00:27:45.929 joaquin: And so, I’m doing, like, the online course hybrid, where I’m taking,
202 00:27:46.060 ⇒ 00:27:52.520 joaquin: I’m starting the online course, and then once the in-person course begins, I’ll be on-site for that.
203 00:27:52.520 ⇒ 00:27:53.080 Uttam Kumaran: Nice.
204 00:27:53.080 ⇒ 00:27:58.270 joaquin: That’s in April. And so, really trying to level up my AI skills, because obviously that’s where it’s heading.
205 00:27:58.270 ⇒ 00:28:04.170 Uttam Kumaran: I’ve worked at a company where, essentially, it was like a block scenario, right? Yeah.
206 00:28:04.520 ⇒ 00:28:17.740 joaquin: we were 80 engineers, high gross, doing $430 million in gross monthly volume through the platform, this is LeafLink, and we contracted from 80 engineers to 8 over the course of 2 years.
207 00:28:17.740 ⇒ 00:28:26.710 joaquin: And so, I want to be kind of part of the future, not, like, behind it. Yeah. Really trying to, double down on my AI knowledge.
208 00:28:26.710 ⇒ 00:28:27.460 Uttam Kumaran: Dope.
209 00:28:27.460 ⇒ 00:28:29.940 joaquin: more recently, I’m trying to…
210 00:28:30.180 ⇒ 00:28:37.489 joaquin: automate… I found two tools that I think that I want to work together, but I just haven’t had the time to,
211 00:28:37.910 ⇒ 00:28:43.970 joaquin: to pursue this as a hobbyist, but, there’s one tool called TD. Let me share that with you.
212 00:28:43.970 ⇒ 00:28:44.590 Uttam Kumaran: Yeah.
213 00:28:45.700 ⇒ 00:28:47.140 joaquin: TD.
214 00:28:49.480 ⇒ 00:28:55.410 joaquin: Which is a task management, for your local AI.
215 00:28:56.070 ⇒ 00:29:01.889 joaquin: And so what I’m trying to use that is essentially pick tickets off of whatever your source is, so, like.
216 00:29:02.180 ⇒ 00:29:06.579 joaquin: Break it down into atomic pieces of work, and then run several agents against it.
217 00:29:06.580 ⇒ 00:29:10.750 Uttam Kumaran: No, I’m doing that… you know what I’m doing today is I, I’m doing that with…
218 00:29:11.070 ⇒ 00:29:15.460 Uttam Kumaran: with Cursor, I’m having… so I have Cursor, basically… we have…
219 00:29:15.530 ⇒ 00:29:27.709 Uttam Kumaran: we have, markdown files on, like, what makes a great Brain Forge ticket, and so I’m able to go… so… so I usually… I’ll spend a bunch of time writing out a big architecture plan. Like, for example, I’m my… I’m…
220 00:29:27.710 ⇒ 00:29:35.790 Uttam Kumaran: I’m migrating some stuff into our platform today. I work on the plan, and then I basically have Cursor break it down into a linear project with tickets.
221 00:29:35.880 ⇒ 00:29:52.889 Uttam Kumaran: And then I’ve been… today, I’m literally, like, staring at Cursor Cloud, because I’m just assigning tickets one by one, and then verifying, and it’s working really, really well. So, like, this is exactly the stuff that we’re doing, but again, you’ll… you’ll be sup… maybe surprised, maybe not, is that, like.
222 00:29:53.040 ⇒ 00:29:56.750 Uttam Kumaran: Companies are just not… they’re, like, just discovering, like.
223 00:29:57.230 ⇒ 00:29:59.320 Uttam Kumaran: Oh, I can use cursor for stuff.
224 00:29:59.320 ⇒ 00:30:00.179 joaquin: Yeah, yeah, for.
225 00:30:00.180 ⇒ 00:30:07.839 Uttam Kumaran: So, it’s… it’s… on one hand, I think there’s gonna be an opportunity for us to do so much with fewer people.
226 00:30:08.080 ⇒ 00:30:25.480 Uttam Kumaran: Not fewer people than now, like, net-net, like, they benchmark fewer people. We need every people, every person that was working here is, like, really, really busy right now. And then, second is, like, I want to basically pay people more, and I want to have fewer people, I want to pay people more, I want us to be on the cutting edge.
227 00:30:25.500 ⇒ 00:30:32.960 Uttam Kumaran: of what it takes to run an AI-native consultancy, and also, like, do AI service, right? It’s like the two-for-one.
228 00:30:32.990 ⇒ 00:30:42.030 Uttam Kumaran: And so, we’re thinking about a lot of the same thing. I mean, I think… I think what you’re saying and the way you’re thinking about it makes a lot of sense, but, like, you’ve done backend for a while, like…
229 00:30:42.070 ⇒ 00:30:47.340 Uttam Kumaran: It’s not that… it’s… I think, where… there’s just a lot of new… concepts.
230 00:30:47.590 ⇒ 00:31:01.579 Uttam Kumaran: But when you… when you go, like, I think when you take the course, you’ll realize that they rhyme a lot with, like, different abstractions and functions and endpoints. In fact, I think the normal… the normal stuff is actually still…
231 00:31:01.770 ⇒ 00:31:03.619 Uttam Kumaran: Like, knowing how to write
232 00:31:03.740 ⇒ 00:31:14.269 Uttam Kumaran: and give good requirements for endpoints, knowing how to debug and knowing, like, what are the common pitfalls, choosing the right architecture still continues to be the most alpha thing. Like.
233 00:31:14.440 ⇒ 00:31:27.729 Uttam Kumaran: Yeah. Because you can run to 100 agents, but if you give it a shitty prompt, you don’t give enough instruction, it’s gonna mess up, and so that’s what we’re finding, is, like, the people on our team know how to build stuff, like, they’re not, like, AI first, meaning…
234 00:31:27.860 ⇒ 00:31:37.239 Uttam Kumaran: we still do what great engineering does, like, you try to write a PRD, you try to write a TDD, that goes through a view. And then the execution is actually what we’ve, like.
235 00:31:37.260 ⇒ 00:31:52.979 Uttam Kumaran: been able to hand off, but… but the execution is so dependent on the quality of that, and so more of our time is gonna shift upwards to, like, actually, we can write a TDD for, like, almost everything, and we can write a PRD, right? The PMs can actually, like.
236 00:31:53.120 ⇒ 00:32:07.969 Uttam Kumaran: demo something just locally as a proof of concept before handing it to, like, an engineering team. So that’s a lot of, like, what we’re thinking about. So, I mean, dude, if, like, that’s what, like, you’re interested in, I feel like there’s totally opportunities here to collaborate,
237 00:32:08.100 ⇒ 00:32:21.019 Uttam Kumaran: I mean, I would love for you to chat with, Sam, who leads, sort of our AI sort of platform team, and just, like, say hi to him and hear more. He’ll go a little bit more in-depth on some of the projects we’re working on, but, like.
238 00:32:21.170 ⇒ 00:32:38.729 Uttam Kumaran: I don’t know, I feel like I really like your attitude, and I think you’re thinking about it kind of where I was a few years ago, where I was like, okay, I think as an engineer, really continue to learn the fundamentals, but then leveraging this to just basically get to, like, that 5, 10, 15x leverage is, like, where this is gonna go.
239 00:32:38.790 ⇒ 00:32:41.270 Uttam Kumaran: And I think we’re in a moment right now that, like.
240 00:32:41.530 ⇒ 00:32:44.749 Uttam Kumaran: Being able to be in an environment that, like, encourages that.
241 00:32:44.830 ⇒ 00:33:02.990 Uttam Kumaran: Versus, like, oh, we don’t… we’re… Cursor is going through, like, IT review, or, like, some BS like that. Like, this is not the case here. It’s like, whatever the opposite of that is, is what the situation is here. It’s so… I’m so annoying about… about this stuff, you know? So…
242 00:33:03.630 ⇒ 00:33:04.939 joaquin: No, that’s good, that’s good.
243 00:33:04.940 ⇒ 00:33:05.610 Uttam Kumaran: Yeah.
244 00:33:05.890 ⇒ 00:33:09.619 joaquin: Okay, Sam is the, the lead, engineer over at…
245 00:33:09.620 ⇒ 00:33:11.420 Uttam Kumaran: Yeah, on the AI side.
246 00:33:11.420 ⇒ 00:33:11.970 joaquin: Okay.
247 00:33:12.520 ⇒ 00:33:17.930 joaquin: Cool. Yeah, I’d be happy to meet with Sam and, and… Understand better, so…
248 00:33:17.930 ⇒ 00:33:37.799 Uttam Kumaran: Okay, okay, perfect. And then, yeah, I mean, also, like, I’m happy to introduce you to Kayla, who leads, like, kind of recruiting and people for us, too, so she can kind of give you a sense, but, like, again, I think chat with Sam and see, like, if you’d be interested in the process. We have openings for more people on our AI team. Again, I think what you can see is, like, we’re doing good practical AI work, so…
249 00:33:37.870 ⇒ 00:33:45.570 Uttam Kumaran: it’s beyond… it’s beyond N8N, it’s beyond just, like, hooking up Zapiers or, like, building mini MCPs, it’s like…
250 00:33:45.610 ⇒ 00:34:03.040 Uttam Kumaran: it’s still… it’s like application development work, except everything has, like, a flavor of AI, where it’s, like, some agentic workflow, or it’s a chat interface, or it’s still a… it’s some type of UI mix of all that, but a lot of it is still the same, like, there needs to be backend, we need to do auth.
251 00:34:03.360 ⇒ 00:34:14.280 Uttam Kumaran: you know, there’s some complexities. And then the fun of it is because we get to use the latest AI, it unlocks new capabilities for our clients, you know? So, in that way, I think the stuff we’re doing is…
252 00:34:14.489 ⇒ 00:34:20.109 Uttam Kumaran: a lot more practical than just, like, your, like, a random automation shop that’s like, oh, I’m gonna automate, like, this into that, like…
253 00:34:20.310 ⇒ 00:34:25.710 Uttam Kumaran: that’s, like, not what we’re doing. It’s more of, like, hey, we walk into a company, and…
254 00:34:25.909 ⇒ 00:34:28.380 Uttam Kumaran: Currently, their customer service reps are, like.
255 00:34:28.870 ⇒ 00:34:32.870 Uttam Kumaran: Ctrl-Fing across, like, 100 documents while they’re on the call.
256 00:34:33.060 ⇒ 00:34:48.839 Uttam Kumaran: And my job isn’t, like, oh, we’re gonna replace… it’s like, they’re not trying to replace anybody. They’re actually just trying to service their customers better, and so let’s build them a chatbot that, like, rags through a bunch of these docs, surfaces the right pieces, like, that’s what a lot of, like, what we’re doing, you know?
257 00:34:49.150 ⇒ 00:35:01.189 joaquin: Okay, cool, cool. And, would you guys ever see, like, an evolution towards, like, building things more from scratch, like custom LLM models, and… or, the…
258 00:35:01.190 ⇒ 00:35:16.950 Uttam Kumaran: Yeah, I hope so, but I feel like… and let’s… that’s more of, like, the engineer in me, like, in terms of the business use case, there is a lot of alpha still in just the regular, regular stuff we’re doing, but yes, like, I want us to continue to
259 00:35:17.150 ⇒ 00:35:19.350 Uttam Kumaran: Build a gap between the next
260 00:35:19.540 ⇒ 00:35:25.169 Uttam Kumaran: like, for me, I think of… when I think about us, I’m like, okay, where is Accenture, or where is Deloitte, and like…
261 00:35:25.250 ⇒ 00:35:40.509 Uttam Kumaran: in what way are we… are we better? Right? And so, for me, I do want to get into fine-tuning. I do want to start having stronger opinions about use case-specific models. Like, what’s better at doing text-to-SQL versus what’s better at doing extractions.
262 00:35:40.560 ⇒ 00:35:59.480 Uttam Kumaran: I want to get more into, like, OCR PDF extraction, and then it would be lovely just to get into, like, model training and deeper into that direction, for sure. So it’s only a matter of time when someone asks us for it. At that point, we will decide whether or not to do it.
263 00:35:59.480 ⇒ 00:36:02.370 joaquin: But, like… They’re just so behind, they don’t know they want to ask for that.
264 00:36:02.370 ⇒ 00:36:15.190 Uttam Kumaran: You’re right, like, that’s the painful part, dude, because last year, when I was pitching AI stuff, nobody really understood. They were, like, still, like, oh, but opening, like, I can do this in ChatGPT, how is this different than ChatGPT?
265 00:36:15.350 ⇒ 00:36:27.450 Uttam Kumaran: I’m like, oh, yeah, like, alright, just, I’ll call you back in, like, 6 months. And so now it looks very, very different, but I do think that we have some type of lead right now that we’re just trying to take advantage of.
266 00:36:27.540 ⇒ 00:36:37.219 Uttam Kumaran: And then continuing… for me, again, the biggest part is just, like, how do I build a sick team? How do I build a dope team of people that we can walk into the next environment?
267 00:36:37.500 ⇒ 00:36:53.710 Uttam Kumaran: even if… even if in 6 months, it’s like, everybody, we have to learn what fine-tuning is, how to do it. Not, like, that worried, right? Really, and the client is actually less worried about us being able to do that. They’re like, these guys can roll with the punches. Like, we don’t want to have to find another
268 00:36:54.150 ⇒ 00:36:57.399 Uttam Kumaran: like, consultancy to work with, right? So that’s a lot of what we’re doing.
269 00:36:58.070 ⇒ 00:36:59.040 Uttam Kumaran: Very cool.
270 00:36:59.040 ⇒ 00:37:00.770 joaquin: Awesome. Exciting time for you.
271 00:37:00.770 ⇒ 00:37:02.270 Uttam Kumaran: Yeah, appreciate it, dude.
272 00:37:03.010 ⇒ 00:37:07.590 Uttam Kumaran: Cool. So let me, let me just, like, connect you with some folks over email.
273 00:37:07.630 ⇒ 00:37:11.909 joaquin: But again, I know we were, like, trying to connect for a while, so thank you for the.
274 00:37:11.910 ⇒ 00:37:12.960 Uttam Kumaran: Thank you for the patience.
275 00:37:12.960 ⇒ 00:37:18.470 joaquin: Yeah, for sure, for sure. I mean, I know you’re busy, and yeah, I was able to keep preoccupied, so…
276 00:37:18.470 ⇒ 00:37:19.490 Uttam Kumaran: Okay, cool.
277 00:37:19.700 ⇒ 00:37:21.410 joaquin: Glad we could connect, so…
278 00:37:21.410 ⇒ 00:37:32.869 Uttam Kumaran: Definitely, yeah, and please feel free to, like, email me if you have, like, any questions or anything, like, I’m really, really an open book, so, like, even if after this conversation you have some questions, like, more than happy to answer.
279 00:37:33.240 ⇒ 00:37:35.130 joaquin: Sick. Bhutan, thank you so much.
280 00:37:35.130 ⇒ 00:37:37.929 Uttam Kumaran: Yeah. Thank you, dude. Okay, I’ll talk to you soon.
281 00:37:38.370 ⇒ 00:37:39.060 Uttam Kumaran: Bye.