Meeting Title: Brainforge x Caleb Data Integration Check-in Date: 2026-01-12 Meeting participants: Caleb, Uttam Kumaran
WEBVTT
1 00:01:11.800 ⇒ 00:01:12.610 Uttam Kumaran: Hey!
2 00:01:14.560 ⇒ 00:01:15.360 Caleb: Hello.
3 00:01:15.630 ⇒ 00:01:16.829 Uttam Kumaran: Hey, how’s it going?
4 00:01:16.830 ⇒ 00:01:17.500 Caleb: Fair enough.
5 00:01:17.880 ⇒ 00:01:18.800 Uttam Kumaran: Good to meet you.
6 00:01:19.850 ⇒ 00:01:20.739 Caleb: Where are you calling from?
7 00:01:21.080 ⇒ 00:01:22.609 Uttam Kumaran: I’m in Austin.
8 00:01:22.610 ⇒ 00:01:23.670 Caleb: Oh, nice.
9 00:01:24.020 ⇒ 00:01:24.350 Uttam Kumaran: Yeah, huh.
10 00:01:25.110 ⇒ 00:01:28.610 Caleb: What is that R? What is the R on the couch? Is that your college?
11 00:01:28.990 ⇒ 00:01:33.700 Uttam Kumaran: It’s a B, yeah, I went to Bucknell. It’s a college in Central PA.
12 00:01:33.930 ⇒ 00:01:35.029 Caleb: Oh, okay, cool.
13 00:01:35.030 ⇒ 00:01:36.109 Uttam Kumaran: Yeah, yeah.
14 00:01:36.670 ⇒ 00:01:41.359 Uttam Kumaran: How did you, how’d you get connected with Sawyer? He’s an old friend of mine from Austin, so…
15 00:01:41.860 ⇒ 00:01:44.820 Caleb: Yeah, we’re basically ex-coworkers from Icon.
16 00:01:45.040 ⇒ 00:01:51.040 Uttam Kumaran: Oh, nice, okay. How was the, how was the whole icon situation?
17 00:01:51.840 ⇒ 00:01:52.820 Caleb: is cooked.
18 00:01:54.280 ⇒ 00:01:59.140 Caleb: Yeah, it was kind of… Yeah, it was a whole ordeal, but we’re out of it now.
19 00:01:59.830 ⇒ 00:02:07.440 Uttam Kumaran: what did you, like, upon reflection, like, what do you, what do you think about the company, or, like, what do you think about that, that industry, or yeah, I’m just curious.
20 00:02:07.810 ⇒ 00:02:13.440 Caleb: I think… Yeah, like… There was something, but we just…
21 00:02:14.340 ⇒ 00:02:21.130 Caleb: Due to bad leadership, we just weren’t… able to… execute properly.
22 00:02:21.310 ⇒ 00:02:22.010 Uttam Kumaran: Yeah.
23 00:02:22.350 ⇒ 00:02:24.220 Caleb: Or execute on the right things.
24 00:02:24.440 ⇒ 00:02:25.640 Uttam Kumaran: Yeah, makes sense.
25 00:02:26.090 ⇒ 00:02:26.620 Caleb: Yeah.
26 00:02:27.140 ⇒ 00:02:34.659 Uttam Kumaran: Well, nice, yeah, I’m happy to… happy to see how we can be helpful. So, again, my name is Utam, I run Brainforge, we’re a data and AI consultancy.
27 00:02:34.930 ⇒ 00:02:43.489 Uttam Kumaran: We do a lot… my background’s in data engineering, do a lot of work in data for folks, so standing up data infrastructure.
28 00:02:43.700 ⇒ 00:03:03.560 Uttam Kumaran: Everything from data warehousing, modeling, sort of, like, as well as a lot on the product analytics side, as well as BI and sort of strategy work, so kind of like full-stack data. And then we also do, you know, a lot of AI and application development work. So probably around, like, 25 people on the team right now.
29 00:03:04.960 ⇒ 00:03:17.770 Uttam Kumaran: And it’s a company I started a few years ago, so we’re just… we’re just growing, and yeah, Sawyer mentioned a little bit about, sort of, like, what you guys are… are thinking about or planning, so yeah, happy to hear more and see if, like, there’s a way
30 00:03:17.880 ⇒ 00:03:19.320 Uttam Kumaran: You know, we could be helpful.
31 00:03:19.670 ⇒ 00:03:25.899 Caleb: Yeah, basically we need to integrate all these connectors and, like, sync all this data into our system, and right now I’m using Airbyte.
32 00:03:26.030 ⇒ 00:03:30.150 Caleb: to… connect to, like, Facebook, meta.
33 00:03:31.990 ⇒ 00:03:36.750 Caleb: I’m trying to figure out the flow right now, but it’s basically, like, dumping everything into Click House.
34 00:03:38.190 ⇒ 00:03:38.740 Uttam Kumaran: Okay.
35 00:03:38.740 ⇒ 00:03:43.040 Caleb: And then from there, We query on the data.
36 00:03:43.040 ⇒ 00:03:43.570 Uttam Kumaran: Okay.
37 00:03:43.620 ⇒ 00:03:45.310 Caleb: In our application.
38 00:03:45.660 ⇒ 00:03:46.160 Uttam Kumaran: Okay.
39 00:03:46.160 ⇒ 00:03:49.639 Caleb: Like, use AI to generate, like, SQL queries and stuff.
40 00:03:49.880 ⇒ 00:03:54.440 Uttam Kumaran: Yeah. Okay. Yeah, familiar. I mean, we’ve done a lot of work on ClickHouse, we do a lot of…
41 00:03:54.630 ⇒ 00:03:56.590 Uttam Kumaran: text-to-SQL work as well.
42 00:03:56.940 ⇒ 00:04:01.930 Uttam Kumaran: You know, we have a couple clients that we’re doing Texas SQL, like, agent work with, and…
43 00:04:02.250 ⇒ 00:04:12.419 Uttam Kumaran: Yeah. So, I guess, like, do you guys have a sense of, sort of, timeline? Like, do you have a few sources that are, like, more priority, or, like, give me a sense of, kind of, like, where the project’s at?
44 00:04:12.750 ⇒ 00:04:16.670 Caleb: Yeah, so, like… I think there…
45 00:04:17.200 ⇒ 00:04:20.980 Caleb: So the meta stuff is, like, pretty much done already.
46 00:04:21.890 ⇒ 00:04:29.509 Caleb: And then, from there, we do, like, reviews, integrating reviews, which the… we hired some interns who are already working on that stuff.
47 00:04:30.370 ⇒ 00:04:34.629 Caleb: And then from there, it’s like… You start to do…
48 00:04:35.900 ⇒ 00:04:40.969 Caleb: Honestly, it’s whatever our customers are… Want us to prioritize.
49 00:04:40.970 ⇒ 00:04:41.530 Uttam Kumaran: Okay.
50 00:04:41.530 ⇒ 00:04:47.179 Caleb: But… Potentially, like, Shopify is, like, the next big one.
51 00:04:47.410 ⇒ 00:04:47.980 Uttam Kumaran: Okay.
52 00:04:48.410 ⇒ 00:04:57.290 Caleb: Yeah, and… I guess, like, something I have to figure out is… if…
53 00:04:58.150 ⇒ 00:05:03.070 Caleb: Like, if you guys do all the, like, grunt work of, like, hooking up pipelines and stuff.
54 00:05:03.360 ⇒ 00:05:03.850 Uttam Kumaran: Fair.
55 00:05:04.070 ⇒ 00:05:09.350 Caleb: I guess that would mean I would be doing, like, other product stuff.
56 00:05:10.200 ⇒ 00:05:12.389 Caleb: I could see that being helpful.
57 00:05:13.370 ⇒ 00:05:19.660 Caleb: I guess, like, Yeah, the other big part is, like.
58 00:05:23.590 ⇒ 00:05:26.859 Caleb: The schemas are stuff that is, like, very…
59 00:05:28.090 ⇒ 00:05:32.769 Caleb: Like, we have to learn them ourselves, regardless of whether or not
60 00:05:33.610 ⇒ 00:05:35.480 Caleb: We bring in, like, you guys.
61 00:05:35.680 ⇒ 00:05:36.190 Uttam Kumaran: Sure.
62 00:05:36.680 ⇒ 00:05:40.899 Caleb: it’s not… I don’t think it’s just like, oh, you guys have hooked it up, now it’s like.
63 00:05:41.510 ⇒ 00:05:44.829 Caleb: Like, we still have to do the same amount of…
64 00:05:45.200 ⇒ 00:05:48.200 Caleb: Like, discovery on it, if that makes sense.
65 00:05:48.530 ⇒ 00:06:04.099 Uttam Kumaran: Yeah, I mean, so all sort of the sources that you mentioned, you know, we do a lot of work in e-com and B2B SaaS and, like, B2C SaaS, and so we’re very familiar with all of the, like, most common, like, e-com sources, so Shopify, Amazon, Walmart.
66 00:06:04.100 ⇒ 00:06:07.000 Caleb: All the ad sources, so Meta…
67 00:06:07.100 ⇒ 00:06:25.699 Uttam Kumaran: Snapchat, TikTok, whatever, as well as, like, you know, most of the most common, like, B2B sources, so all product analytics, like Amplitude, Mixpanel, Shop, like, Stripe, you know, things like that. So, yeah, we… we have a lot of familiar with those schemas, so landing that data and putting it into a place where it is queryable,
68 00:06:25.700 ⇒ 00:06:35.309 Uttam Kumaran: is not really a problem for us, but certainly, like, we can… we can help on, you know… my background’s also in product. I led product at a…
69 00:06:35.390 ⇒ 00:06:46.179 Uttam Kumaran: Data startup, like, before this. So, more than happy to collaborate on if you need help on, like, you know, how… if we can also help to speed up stuff on, like, sort of the product design aspect.
70 00:06:46.180 ⇒ 00:06:50.570 Caleb: The schemas for all of these are… the more common the source.
71 00:06:50.700 ⇒ 00:06:57.230 Uttam Kumaran: It’s kind of easier to… to kind of get ahead on the scheme and understand, like, what… what you’ll have on the… on the… basically to use for the front end, right?
72 00:06:57.750 ⇒ 00:07:01.780 Caleb: I guess, like, we already, like, I already set up this flow for, like.
73 00:07:02.060 ⇒ 00:07:06.679 Caleb: integrating with AirByte, and AirByte obviously provides a ton of
74 00:07:06.980 ⇒ 00:07:12.040 Caleb: Integrations already, which should cover us for, like, most of the stuff already.
75 00:07:12.180 ⇒ 00:07:12.750 Uttam Kumaran: Yeah.
76 00:07:12.950 ⇒ 00:07:15.950 Caleb: So I guess, like, the data…
77 00:07:16.110 ⇒ 00:07:19.209 Caleb: The basic initial data pipeline is already built out.
78 00:07:19.530 ⇒ 00:07:19.930 Uttam Kumaran: Okay.
79 00:07:19.930 ⇒ 00:07:30.200 Caleb: But, again, like, since I’ve personally never done, like, data engineering, like, I’m not sure, like…
80 00:07:31.130 ⇒ 00:07:32.609 Caleb: If there’s gonna be a…
81 00:07:32.750 ⇒ 00:07:37.139 Caleb: if there will come time in the near future where it’s like, I don’t know…
82 00:07:37.350 ⇒ 00:07:39.150 Caleb: Like, I don’t know what I don’t know, you know?
83 00:07:39.150 ⇒ 00:07:40.240 Uttam Kumaran: Yeah, yeah.
84 00:07:40.240 ⇒ 00:07:43.610 Caleb: I don’t even know if, like, there’s gonna be…
85 00:07:44.010 ⇒ 00:07:46.290 Caleb: A problem with, like, our current setup, or…
86 00:07:46.290 ⇒ 00:07:46.950 Uttam Kumaran: Sure.
87 00:07:46.950 ⇒ 00:07:48.060 Caleb: But, yeah.
88 00:07:48.060 ⇒ 00:07:59.289 Uttam Kumaran: Yeah, another good way to leverage us is, like, even if it’s not the right time, like, we could still help, like, probably on an hourly basis, just if you’re like, hey, I… can you just, like, gut check this architecture?
89 00:07:59.290 ⇒ 00:07:59.790 Caleb: Yeah.
90 00:07:59.790 ⇒ 00:08:04.379 Uttam Kumaran: we’ve done a lot of work with all the most common ETL sources, but, like, moving data from
91 00:08:04.760 ⇒ 00:08:15.140 Uttam Kumaran: endpoints, whatever the format, into structured, like, data warehousing or click house. That’s, like, all… that’s, like, basically all we do every day for a bunch of people, so even if you’re like, hey.
92 00:08:15.400 ⇒ 00:08:21.479 Uttam Kumaran: can I just, like, grab an hour to, like, walk through this architecture? Or you’re like, hey, I need help sort of scaling this.
93 00:08:21.480 ⇒ 00:08:21.930 Caleb: Yeah.
94 00:08:21.930 ⇒ 00:08:27.100 Uttam Kumaran: process out. Whether it’s even, like, yeah, whatever that is, like, that could be a good way.
95 00:08:27.310 ⇒ 00:08:30.330 Uttam Kumaran: The leverage us, too, before, like, doing something bigger.
96 00:08:30.860 ⇒ 00:08:37.320 Caleb: Do you guys use, like… Like, I guess you’re… you guys are basically, like,
97 00:08:38.169 ⇒ 00:08:41.929 Caleb: A white glove air bite, almost.
98 00:08:42.490 ⇒ 00:08:45.009 Uttam Kumaran: No, I mean, not real.
99 00:08:45.010 ⇒ 00:08:45.980 Caleb: an offering.
100 00:08:46.620 ⇒ 00:09:01.479 Uttam Kumaran: I mean, we… so we work with, like, a ton of clients, so we’re like a… we’re like a consultancy, so we do… not only do we do data engineering, but it’s also modeling, BI, strategy, like, product analytics, so data engineering is probably, like.
101 00:09:01.690 ⇒ 00:09:06.589 Uttam Kumaran: the smallest thing that we do, just because we do, like, so many… we do a bunch of things for a bunch of clients.
102 00:09:06.620 ⇒ 00:09:07.740 Caleb: So…
103 00:09:07.740 ⇒ 00:09:26.869 Uttam Kumaran: we don’t… we don’t manage applications, we don’t build… like, we don’t have… we don’t, like, white-label anything. Oh, I see. So, we’re… on behalf of clients, we’re helping them make purchasing decisions, whether it’s Fivetran, Polyatomic, Estuary. Most of our clients are not… like, some of our application clients are using Airbite under the hood.
104 00:09:26.870 ⇒ 00:09:32.620 Uttam Kumaran: But most of our, like, e-com or B2B SaaS clients typically just are buying Fivetran. Like, we’re not…
105 00:09:32.630 ⇒ 00:09:38.370 Uttam Kumaran: We’re helping them make the procurement decisions, like, we don’t manage… meaning, like, we’re not .
106 00:09:38.710 ⇒ 00:09:45.549 Caleb: Do you have a… Do you have a… like, Fivetran versus AirByte, like, what are your thoughts there?
107 00:09:46.300 ⇒ 00:09:58.190 Uttam Kumaran: I mean, AirByte is, like, I mean, there’s… it’s an… it’s kind of an open source project, and, like, a lot of them… if you have, like, a really high volume, really, important SLA,
108 00:09:58.510 ⇒ 00:10:10.190 Uttam Kumaran: kind of connector, I… we encourage everybody to usually use Fivetrain or Polytomic, just because you can get support. Like, we’re working with, like, brands that are, like, you know, making a couple hundred million dollars.
109 00:10:10.300 ⇒ 00:10:20.340 Uttam Kumaran: they’re, like, they can’t rely on AirByte, because it’s just, like, an open source thing that’s, like, also very new. But for companies that are building applications really quickly, it’s a great, cheap way to kind of get started.
110 00:10:20.580 ⇒ 00:10:38.550 Uttam Kumaran: You can get the same capabilities, though, out of Fivetran, Polytomic, and others, so, like, if budget isn’t a problem, I would probably encourage y’all to… to look that way, because their support and the quality of the connectors and the breadth are gonna be way higher. Everybody’s just a good way to start, basically.
111 00:10:39.030 ⇒ 00:10:44.580 Caleb: Basically, our use case is we need to… Like, do per tenant. Yeah.
112 00:10:44.610 ⇒ 00:10:45.590 Uttam Kumaran: Right. Yeah.
113 00:10:45.590 ⇒ 00:10:46.050 Caleb: Which…
114 00:10:46.050 ⇒ 00:10:53.330 Uttam Kumaran: You can do it on Fivetran, I guess it just depends on… Fivetrane’s gonna be the most expensive option, but, like, the… probably the best, most,
115 00:10:53.580 ⇒ 00:10:57.550 Uttam Kumaran: Like, the oldest product in this space, so, like.
116 00:10:57.650 ⇒ 00:11:03.410 Uttam Kumaran: They do power… you can do power by Fivetran, so you can basically white-label them, they’ll handle the handoffs.
117 00:11:03.410 ⇒ 00:11:05.920 Caleb: And trigger the syncs to wherever, and you can…
118 00:11:05.920 ⇒ 00:11:14.680 Uttam Kumaran: you can do that basically multi-tenant, you can kind of hand all that off to them, versus the Airbuy thing, you’re gonna have to build on your… kind of on your own. I have another friend that’s…
119 00:11:14.680 ⇒ 00:11:32.219 Uttam Kumaran: They… they are doing, like, an Ecom-based AI agent that… for ECOM data analysis. Happy to connect you with them, they just basically built this, but I think they… they switched from using AirByte to Fivetran recently, just because it’s like… for me, I’m like, look, this is just not a core part of the product, it’s just, like, brokering the data.
120 00:11:32.220 ⇒ 00:11:34.730 Caleb: And so I’m kind of, like, handed off, if you can.
121 00:11:34.860 ⇒ 00:11:38.219 Uttam Kumaran: There’s… but it kind of depends, like, if you’re, like.
122 00:11:38.440 ⇒ 00:11:44.219 Uttam Kumaran: Okay. Yeah. Just again, like, managing… managing these pipelines, it’s just, like, not something, like…
123 00:11:44.650 ⇒ 00:11:51.279 Uttam Kumaran: What’s core to the product is that it works, and you kind of, like, get it to sync, and you kind of, like, hand it off, is my perspective.
124 00:11:51.280 ⇒ 00:11:58.120 Caleb: Yeah. Okay, so, I guess for AirByte, like, we’re using their cloud offering.
125 00:11:58.120 ⇒ 00:11:58.710 Uttam Kumaran: Yeah.
126 00:11:58.750 ⇒ 00:12:03.660 Caleb: And… the… how… This is what I’m trying to figure out right now.
127 00:12:03.660 ⇒ 00:12:08.419 Uttam Kumaran: Are you putting the… are you white-labeling it, or you’re just getting the keys from people and, like, kind of doing it on the back end?
128 00:12:08.420 ⇒ 00:12:09.889 Caleb: Oh, white… we’re white-labeling it.
129 00:12:09.890 ⇒ 00:12:10.790 Uttam Kumaran: Okay, cool.
130 00:12:10.790 ⇒ 00:12:13.189 Caleb: Yeah, so they have, like, an embedded thing.
131 00:12:13.190 ⇒ 00:12:14.670 Uttam Kumaran: Cool. Yeah.
132 00:12:14.690 ⇒ 00:12:22.709 Caleb: Right, okay, and then, the… Right now, it creates, like, One table per tenant.
133 00:12:23.020 ⇒ 00:12:25.760 Caleb: Yeah.
134 00:12:25.760 ⇒ 00:12:29.579 Uttam Kumaran: It’s just… let’s just take a Shopify, like, for what connector? Like, Shopify? Or for Meta?
135 00:12:29.580 ⇒ 00:12:30.960 Caleb: For me, for me.
136 00:12:30.960 ⇒ 00:12:31.570 Uttam Kumaran: Yeah.
137 00:12:31.570 ⇒ 00:12:36.729 Caleb: So I’m basically trying to, like, sync them into just one table instead.
138 00:12:36.730 ⇒ 00:12:39.320 Uttam Kumaran: Yeah, so a couple things you’ll be able to do. One is, like.
139 00:12:39.830 ⇒ 00:12:54.920 Uttam Kumaran: you can trigger all of those, so, like, it’ll land, but then you probably need, like, something like dbt on top of it, which what’s gonna do is gonna run a select and union, and then you can union all that together. Yeah. If that’s sort of, like, what you’re doing. That’s, like, the most common…
140 00:12:55.450 ⇒ 00:13:01.220 Caleb: What they recommended was dumping it into S3 and then going from S3 to Click House.
141 00:13:01.350 ⇒ 00:13:04.009 Uttam Kumaran: You could also do that, yeah. It sort of depends.
142 00:13:04.010 ⇒ 00:13:07.640 Caleb: I would suggest, like, if it’s… if you’re gonna have multiple connectors.
143 00:13:07.640 ⇒ 00:13:11.529 Uttam Kumaran: then, yeah, it may be easier just to just dump it as S3 as, like, parquet.
144 00:13:12.500 ⇒ 00:13:20.759 Uttam Kumaran: folder for every client, and then stream all of that into ClickHouse, like, using, like, copy. Basically, like, it’ll… they have a native connector, I think.
145 00:13:22.240 ⇒ 00:13:31.600 Uttam Kumaran: And then, yeah, you can sort of run, but then you’ll still be… have to, like… it depends. If you’re… if you’re gonna run something on top of all the client data, then there’s a different use case than, like.
146 00:13:31.860 ⇒ 00:13:42.580 Uttam Kumaran: just one by one, but yeah, you can basically stream all that to, like, have all that land into a data lake, and then ClickHouse will natively, like, listen and pick it up and ingest it in.
147 00:13:42.940 ⇒ 00:13:46.720 Uttam Kumaran: It sort of also depends on, like, what your SLAs are for, like…
148 00:13:46.720 ⇒ 00:13:48.620 Caleb: I don’t think our SLAs are, like…
149 00:13:48.870 ⇒ 00:13:56.879 Uttam Kumaran: Yeah, so then… Yeah, then, like, landing into S3 is a good thing, because also you may sh… you may, like, ditch Click House later.
150 00:13:57.260 ⇒ 00:14:02.070 Uttam Kumaran: And you sort of don’t want to have to… True. …deal with that. So, I would… yeah, that’s a…
151 00:14:02.210 ⇒ 00:14:13.989 Uttam Kumaran: pretty good way. So, Airbyte’s cloud offering, if you’re already going with their cloud offering, like, there’s… I would suggest considering two more vendors, just to even try to do a POC. There’s a company called Polytomic. We actually switched the
152 00:14:14.110 ⇒ 00:14:24.199 Uttam Kumaran: all of our ingestion to going through Polyatomic. There’s a smaller team, we got much better support. I’ve used Fivetran for, like, almost, like, 10 years now, so…
153 00:14:24.200 ⇒ 00:14:27.630 Caleb: They’re just support kind of got worse over time, and they’re, like, very expensive.
154 00:14:27.760 ⇒ 00:14:33.630 Uttam Kumaran: So Polytomic is kind of in the middle, but, like, way better support. Polytomic.com, like, P-O-L-Y…
155 00:14:34.020 ⇒ 00:14:36.959 Caleb: But what is their, like, embedded use case, or their…
156 00:14:36.960 ⇒ 00:14:45.839 Uttam Kumaran: Their whole thing, their whole product, you can actually run everywhere from, like, anywhere from on-prem to using their cloud offering. So they have, like, an embed…
157 00:14:46.350 ⇒ 00:14:48.579 Uttam Kumaran: Type product that you can basically white label.
158 00:14:48.820 ⇒ 00:14:53.910 Uttam Kumaran: And, like, the whole thing is very programmable. The thing is, for my business, like.
159 00:14:54.100 ⇒ 00:15:08.769 Uttam Kumaran: since I’m not, like, white-labeling them, I actually just need them to work all the time and have great support, because we’re, like, client… we’re doing consulting. So it’s actually much more important for me to, like, have them on the line in case something goes down for them to work with the client.
160 00:15:09.140 ⇒ 00:15:12.990 Uttam Kumaran: But I would suggest trying Fivetran, or trying,
161 00:15:13.770 ⇒ 00:15:17.720 Uttam Kumaran: Or trying them. Fivetran will also probably help you build it out, too.
162 00:15:19.260 ⇒ 00:15:28.860 Uttam Kumaran: Both of these guys, when you go engage with them, you should just try to have them build a POCS so you don’t have to spend time doing it. Fivetran, they have a lot of people around that are, like, powered by Fivetran product.
163 00:15:29.120 ⇒ 00:15:29.760 Caleb: Yeah.
164 00:15:29.760 ⇒ 00:15:33.649 Uttam Kumaran: They’re, like, sales engineers will help you build it, the first version.
165 00:15:33.950 ⇒ 00:15:34.510 Caleb: Oh, really?
166 00:15:34.510 ⇒ 00:15:42.240 Uttam Kumaran: Yeah, so when you talk to Fivetran especially, and for both these, like, have them try to build, like, build out some of it for you, so you don’t have to, like, do it yourself.
167 00:15:42.840 ⇒ 00:15:47.369 Uttam Kumaran: 5chain, this is a huge business line for them, is, like, kind of, like, white-labeled connectors.
168 00:15:48.140 ⇒ 00:15:52.220 Uttam Kumaran: And you can customize everything, like the look and feel and stuff, so…
169 00:15:52.220 ⇒ 00:16:01.789 Caleb: How much is Fivetran? Because I know for Airbyte’s pricing, they told… they quoted us, like, $4 per tenant, unlimited syncing.
170 00:16:01.790 ⇒ 00:16:04.529 Uttam Kumaran: Yeah, I think Five Train is gonna be…
171 00:16:05.120 ⇒ 00:16:11.409 Uttam Kumaran: It’s gonna be based on rows, and they will sort of, like, it’ll sort of be bucketed, meaning, like.
172 00:16:11.720 ⇒ 00:16:23.800 Uttam Kumaran: Like, as you start to ingest a lot of people’s stuff, you can either decide to pass the cost on, or take it on, but they scale… it’s, like, all, like, monthly active rows.
173 00:16:24.150 ⇒ 00:16:28.300 Uttam Kumaran: So, it’s gonna… it’s definitely gonna be more expensive than AirByte, for sure.
174 00:16:28.450 ⇒ 00:16:33.219 Uttam Kumaran: But it sort of depends on your model. Like, if you’re gonna charge a customer based on…
175 00:16:33.220 ⇒ 00:16:39.980 Caleb: We want to charge the customer, like, $50 a month. Very, like, consumer… like, we want to go for…
176 00:16:39.980 ⇒ 00:16:42.330 Uttam Kumaran: Yeah, yeah, yeah. I think, like, it depend… I mean…
177 00:16:42.500 ⇒ 00:16:47.530 Uttam Kumaran: Yeah, I think it’s gonna depend on what, but also it’s gonna depend on a couple things. One is, like, for Meta.
178 00:16:47.680 ⇒ 00:16:58.770 Uttam Kumaran: there’s, like, tables on Meta that are, like, hourly tables. So, like, it depends, like, what you’re bringing, also, from some sources. Like, if you’re not bringing… for example, Meta, you’ll have, like, campaign ads, ad accounts.
179 00:16:58.770 ⇒ 00:16:59.220 Caleb: Yeah.
180 00:16:59.220 ⇒ 00:17:11.280 Uttam Kumaran: like, ad sets, and, like, they have reports. So, it kind of depends on what you’re bringing in. So, kind of per ingestion, again, when you call Fivetran, they have… they actually have, like, a calculator on their site.
181 00:17:11.470 ⇒ 00:17:18.460 Uttam Kumaran: they’ll go through all the pricing with you on, like, hey, give us, like, an estimate, like, what it is.
182 00:17:18.790 ⇒ 00:17:21.329 Uttam Kumaran: And so, but again, 5chan’s, like, the Cadillac option.
183 00:17:21.660 ⇒ 00:17:24.830 Uttam Kumaran: So, like, I would talk to them just to get a sense of, like, what
184 00:17:25.119 ⇒ 00:17:33.820 Uttam Kumaran: they’re, like, really, like, just probably the best, most stable. And then probably AirByte is, like, the jankiest, newest, cheapest.
185 00:17:34.010 ⇒ 00:17:38.250 Uttam Kumaran: In my experience. And then Polytomic is kind of in the middle. Again, for us, like.
186 00:17:38.690 ⇒ 00:17:43.319 Uttam Kumaran: my client’s paying for everything, so I’m like, I just want the thing that works the best and is the most stable.
187 00:17:43.420 ⇒ 00:18:01.759 Uttam Kumaran: And Polyatomic and Fivetran, the capabilities are so close. In fact, we went with Polyatomic because they help us build new connectors. Like, we go to clients, and they’re like, have this random endpoint, and we’re like, okay, Polyatomic will build it for us, and build it for them, versus Fivetran, they just kind of have a breadth of things.
188 00:18:02.230 ⇒ 00:18:04.909 Uttam Kumaran: So I would suggest
189 00:18:05.370 ⇒ 00:18:11.909 Uttam Kumaran: talking to both of them and kind of seeing, like, what their pricing is, and seeing if they would work with you to build, like, the POC.
190 00:18:13.080 ⇒ 00:18:26.209 Uttam Kumaran: Got it. Yeah. Okay. And they help… they both have all the docs on, like, what the schemas are for what they’ll land, so it does kind of depend on, like, what data you guys are looking for. For example, some of your connectors, you may need to, like, configure reports.
191 00:18:26.230 ⇒ 00:18:28.439 Caleb: Versus some connectors, you’re just gonna get…
192 00:18:28.970 ⇒ 00:18:30.390 Uttam Kumaran: Like, whatever.
193 00:18:30.390 ⇒ 00:18:31.530 Caleb: Configure reports, what do you mean?
194 00:18:31.530 ⇒ 00:18:34.270 Uttam Kumaran: For example, for, like, Google Ads, right?
195 00:18:34.630 ⇒ 00:18:44.589 Uttam Kumaran: you may want to say, like, I want these metrics and these dimensions, versus just, like, getting whatever they give you. Because there’s, like, different granularities of reports that you’ll need.
196 00:18:44.990 ⇒ 00:18:51.200 Uttam Kumaran: Versus, like, a Shopify, they have, like, 15 tables that they’ll just give you, like, orders, order items…
197 00:18:51.470 ⇒ 00:18:54.139 Uttam Kumaran: You know, a bunch, bunch of those, so…
198 00:18:54.540 ⇒ 00:18:55.640 Caleb: Okay, okay.
199 00:18:57.360 ⇒ 00:19:00.710 Uttam Kumaran: The big thing here is just gonna be very… it’s very, like…
200 00:19:00.860 ⇒ 00:19:04.240 Uttam Kumaran: Vendor-dependent, like, on the integration, on, like, what you’re gonna get.
201 00:19:04.970 ⇒ 00:19:10.409 Uttam Kumaran: But starting out with, like, let’s just attack ads first is a good way of thinking about it, versus then
202 00:19:10.580 ⇒ 00:19:24.109 Uttam Kumaran: don’t, like, nail ads, and so nail Meta, Snapchat, all the ad sources, because then you’ll understand, like, what is your universal schema? Like, what are the things you need from every single source, at minimum, to, like, power the product? And then…
203 00:19:24.620 ⇒ 00:19:27.829 Uttam Kumaran: moving on to Shopify, Amazon, or whatever the next
204 00:19:27.980 ⇒ 00:19:34.210 Uttam Kumaran: sort of a category of vendors are is a good way. Otherwise, you’ll… they’re just… there’s, like, a… they’re just so different.
205 00:19:34.670 ⇒ 00:19:35.449 Uttam Kumaran: You know…
206 00:19:37.580 ⇒ 00:19:38.470 Caleb: Okay.
207 00:19:40.540 ⇒ 00:19:41.770 Caleb: That’s helpful.
208 00:19:42.810 ⇒ 00:19:44.419 Caleb: I guess… hmm.
209 00:19:45.120 ⇒ 00:19:49.670 Caleb: how I was originally building the… like…
210 00:19:51.490 ⇒ 00:20:01.730 Caleb: I was gonna build the tools in a way where it’s, like, each connector, like, meta versus… meta ads versus TikTok ads versus…
211 00:20:01.980 ⇒ 00:20:07.390 Caleb: Facebook versus, like, Snapchat ads would be, like, different,
212 00:20:08.620 ⇒ 00:20:11.910 Caleb: different, like, system prompts entirely, where it’s like…
213 00:20:13.080 ⇒ 00:20:16.420 Caleb: It… like, we do discovery on the schema.
214 00:20:16.620 ⇒ 00:20:19.540 Caleb: And then we write a custom system prompt.
215 00:20:20.950 ⇒ 00:20:31.349 Caleb: Because I’m guessing, like… okay, candidly, I haven’t, like, touched any of their platforms yet, but I’m guessing, like, the notion of, like, campaign or ad set might not be the…
216 00:20:32.050 ⇒ 00:20:35.029 Caleb: Be, like, exist at all on, like, another…
217 00:20:35.790 ⇒ 00:20:38.020 Caleb: You know, like an art platform.
218 00:20:38.670 ⇒ 00:20:42.429 Uttam Kumaran: Yeah, I mean, I think, like, doing a system prompt per…
219 00:20:44.450 ⇒ 00:20:57.839 Uttam Kumaran: Yeah, again, I kind of don’t know a ton about the product, so it kind of depends on the questions that are getting answered, getting asked, but yes, like, you… basically, for the best AI, you just want to provide it with an understanding of the schema.
220 00:20:57.840 ⇒ 00:20:58.850 Caleb: And then…
221 00:20:59.010 ⇒ 00:21:06.929 Uttam Kumaran: And within advertising, or digital advertising, there’s… there’s only, like, there’s an 80-20, like, most of the questions you can guess beforehand.
222 00:21:07.060 ⇒ 00:21:15.600 Uttam Kumaran: And so, yeah, but they’re all gonna… they are gonna require some minimums. For example, most people are gonna say, like, how did my campaign perform yesterday?
223 00:21:15.600 ⇒ 00:21:16.150 Caleb: Exactly.
224 00:21:16.150 ⇒ 00:21:18.940 Uttam Kumaran: And, like, so there is a… there will be… you will see an
225 00:21:19.050 ⇒ 00:21:31.570 Uttam Kumaran: ad, ad account, ad set, campaign, like, object across all of your ads. Similarly, like, how many orders did I get? You’ll see some of that. More of the complexity happens where, like.
226 00:21:31.790 ⇒ 00:21:36.529 Caleb: Shopify may have, like, Shopify unique objects that, like, Amazon doesn’t have.
227 00:21:36.570 ⇒ 00:21:41.139 Uttam Kumaran: And, like, how do you… Do you support that or not?
228 00:21:42.210 ⇒ 00:21:43.960 Uttam Kumaran: You know, is, like, the bigger thing.
229 00:21:44.980 ⇒ 00:21:46.420 Caleb: Okay, that’s super helpful.
230 00:21:47.160 ⇒ 00:21:51.249 Caleb: Do you know anyone in our… who’s doing something similar to what we’re doing?
231 00:21:51.770 ⇒ 00:21:59.070 Uttam Kumaran: Yes, I have a couple friends that are doing something like this. I’m happy to connect you with them, you can chat, like, they… some friends that are building, like, e-commerce.
232 00:21:59.500 ⇒ 00:22:01.800 Uttam Kumaran: AI agents,
233 00:22:01.900 ⇒ 00:22:10.590 Uttam Kumaran: We also do a lot of work with, like, Omni, which is a BI tool. They have, like, an AI agent built into their tool. I can send you a couple.
234 00:22:10.760 ⇒ 00:22:11.900 Caleb: Omni Analytics.
235 00:22:11.900 ⇒ 00:22:12.720 Uttam Kumaran: Yeah.
236 00:22:12.720 ⇒ 00:22:14.659 Caleb: Oh yeah, my friend was telling me about this.
237 00:22:14.870 ⇒ 00:22:17.060 Uttam Kumaran: Yeah, so we do a lot of, like, work with them.
238 00:22:17.310 ⇒ 00:22:24.200 Uttam Kumaran: Yeah. There’s, like… I’ll send you a couple links here. These two are also companies you should check out. Merge.
239 00:22:25.050 ⇒ 00:22:31.320 Uttam Kumaran: maybe, maybe check out Merge, like, I feel like they sort of were built, kind of, like, with this focus in mind.
240 00:22:31.980 ⇒ 00:22:34.650 Uttam Kumaran: Merge.dev, yeah. So, like…
241 00:22:35.170 ⇒ 00:22:40.919 Uttam Kumaran: They kind of, like, tried to push this, like, unified schema, and they have, like, sort of these… these products.
242 00:22:41.170 ⇒ 00:22:46.609 Uttam Kumaran: Could be also worth considering. And then Wabi was another AI tool that we.
243 00:22:47.490 ⇒ 00:22:48.430 Caleb: We’ve…
244 00:22:48.430 ⇒ 00:22:54.910 Uttam Kumaran: basically, just did a proof of concept with recently that was really good.
245 00:22:55.690 ⇒ 00:23:03.270 Uttam Kumaran: And then… This is my friend Clint’s company, Hazel.
246 00:23:05.070 ⇒ 00:23:07.350 Caleb: But they’re purely focused on, like.
247 00:23:07.790 ⇒ 00:23:10.400 Uttam Kumaran: I think they’re just doing Shopify for now.
248 00:23:12.190 ⇒ 00:23:13.560 Caleb: Oh, interesting, okay.
249 00:23:17.420 ⇒ 00:23:19.620 Caleb: I see, okay.
250 00:23:19.620 ⇒ 00:23:21.439 Uttam Kumaran: Yeah. And then I know, like.
251 00:23:22.510 ⇒ 00:23:23.210 Caleb: That’s see.
252 00:23:23.510 ⇒ 00:23:28.100 Uttam Kumaran: These guys at TextQL, they’re doing… Oh, yeah, yeah
253 00:23:28.650 ⇒ 00:23:31.740 Uttam Kumaran: Yeah, like, I used to know the guy there, I haven’t talked to him in a while.
254 00:23:32.250 ⇒ 00:23:33.219 Uttam Kumaran: You have a…
255 00:23:33.220 ⇒ 00:23:34.840 Caleb: It’s there.
256 00:23:35.080 ⇒ 00:23:38.519 Uttam Kumaran: Yeah, so TextQL, and then, yeah, I feel like, for the most part, though.
257 00:23:39.100 ⇒ 00:23:43.290 Uttam Kumaran: we’re using… we’re using Omni for… for a lot, and then…
258 00:23:44.160 ⇒ 00:23:48.230 Uttam Kumaran: We found Wabi… Wabi and Omni to kind of be, like, the best two.
259 00:23:50.930 ⇒ 00:23:52.520 Caleb: Bobby and Tommy.
260 00:23:52.800 ⇒ 00:23:53.460 Uttam Kumaran: Yeah.
261 00:23:55.410 ⇒ 00:24:03.660 Caleb: And then, are you, are you, like, did you actually, like, work with these teams, or…
262 00:24:04.620 ⇒ 00:24:12.650 Uttam Kumaran: With… well, like, so we’re… so we’re, like, I mean, we… we’ll purchase a bunch of these on behalf of clients, so, like.
263 00:24:12.800 ⇒ 00:24:16.420 Uttam Kumaran: Yeah, we do a lot of Omni work. We tested Wabi internally.
264 00:24:16.420 ⇒ 00:24:18.510 Caleb: Oh, like, you’re integrating with those.
265 00:24:19.090 ⇒ 00:24:24.969 Uttam Kumaran: No, I mean, I’m a… I run a consultancy, so we have, like, 15 clients, for which we buy… we’re buying, like.
266 00:24:24.970 ⇒ 00:24:25.530 Caleb: Hundreds.
267 00:24:25.530 ⇒ 00:24:27.800 Uttam Kumaran: software for on behalf of them, you know?
268 00:24:28.170 ⇒ 00:24:29.170 Caleb: Got it, got it.
269 00:24:29.170 ⇒ 00:24:30.429 Uttam Kumaran: We don’t have a product.
270 00:24:30.690 ⇒ 00:24:32.189 Caleb: Yeah, yeah, yeah, I see.
271 00:24:34.920 ⇒ 00:24:40.870 Caleb: Or, no, I mean, like, your… your customers are buying…
272 00:24:40.870 ⇒ 00:24:43.899 Uttam Kumaran: Yeah, they’re… we’re referring them to purchase.
273 00:24:44.210 ⇒ 00:24:44.680 Caleb: God.
274 00:24:44.680 ⇒ 00:24:47.140 Uttam Kumaran: Tools, depending on whatever we recommend, yeah.
275 00:24:47.140 ⇒ 00:24:51.109 Caleb: And your normal customer is, like, an e-com brand, basically.
276 00:24:51.750 ⇒ 00:25:00.680 Uttam Kumaran: Yeah, I mean, we’re working… most of our customers are, like, typically 20 million and up, and then we have a couple of brands that are, like, a couple hundred million in revenue.
277 00:25:01.350 ⇒ 00:25:02.530 Uttam Kumaran: Do you… Huh.
278 00:25:02.530 ⇒ 00:25:05.729 Caleb: So, okay, do you work with any software companies as well?
279 00:25:06.310 ⇒ 00:25:10.320 Uttam Kumaran: Yeah, I mean, we work with a bunch of, like,
280 00:25:10.480 ⇒ 00:25:13.769 Uttam Kumaran: like, we’ve done work, like, Bolt was one of our customers.
281 00:25:13.900 ⇒ 00:25:19.959 Uttam Kumaran: We’ve done… we do… we work with a company called Hydra, they’re, like, an AI company. So we have some startups.
282 00:25:20.190 ⇒ 00:25:25.980 Caleb: But… When you’re working for these startups, it’s like they want to integrate, like.
283 00:25:26.150 ⇒ 00:25:32.220 Caleb: analytics and stuff into not their product, but for internally, I guess.
284 00:25:32.220 ⇒ 00:25:39.070 Uttam Kumaran: We also, we also do that, too. So we also are building, helping product startups build AI applications.
285 00:25:39.750 ⇒ 00:25:45.940 Uttam Kumaran: So we have a few companies for which we’re helping in the actual building, like, we do full-stack work building there.
286 00:25:46.060 ⇒ 00:25:50.050 Uttam Kumaran: Helping build our products. Most… again, but we’re not, like, a dev shop.
287 00:25:50.460 ⇒ 00:26:05.950 Uttam Kumaran: So, like, we don’t… we’re, like, we… we own the rela… we own the relationship directly with, like, the CEOs of those companies, and, like, if they need help building out a couple features, or building out the AI piece of their application, we will help do that, but for the most part.
288 00:26:05.950 ⇒ 00:26:14.329 Uttam Kumaran: for B2B SaaS companies, we’re, like, running their data, so all their internal retention, product analytics, customer analytics.
289 00:26:14.590 ⇒ 00:26:16.950 Uttam Kumaran: Yeah.
290 00:26:17.070 ⇒ 00:26:20.810 Uttam Kumaran: I mean, we are consulting with some people on building data applications.
291 00:26:23.330 ⇒ 00:26:24.090 Caleb: Okay.
292 00:26:24.250 ⇒ 00:26:26.409 Caleb: Cool, that’s… that makes a lot of sense.
293 00:26:28.610 ⇒ 00:26:36.830 Caleb: Yeah, I guess I’m… I’m just, like, Trying to understand, like, what… like… like, wit…
294 00:26:37.280 ⇒ 00:26:41.110 Caleb: Like, what the next steps are for this.
295 00:26:41.440 ⇒ 00:26:55.810 Uttam Kumaran: Yeah, I mean, you kind of tell me, like, where you’d like us to be helpful. I mean, we kind of do… I mean, that’s what I’ll have to, like, think about, yeah. Yeah. Like, even if conversations like this are helpful, we can do some type of hourly engagement where I just, like, here, helping out.
296 00:26:55.810 ⇒ 00:26:58.680 Caleb: Like, that’s a lot of ways that people have leveraged us until…
297 00:26:58.680 ⇒ 00:27:10.139 Uttam Kumaran: something bigger makes sense, happy to do that, because this is all we do, so, like, we know most of the vendors in, like, in the space, and kind of have opinions on the best way to architect things, so… yeah.
298 00:27:10.500 ⇒ 00:27:13.439 Caleb: Yeah, I think that probably makes the most sense right now.
299 00:27:13.440 ⇒ 00:27:14.040 Uttam Kumaran: Yeah.
300 00:27:15.410 ⇒ 00:27:16.360 Caleb: Cool.
301 00:27:18.490 ⇒ 00:27:33.000 Uttam Kumaran: Okay, perfect. Well, yeah, maybe, like, maybe I can follow up just, like, kind of, like, what our pricing would look like on that, and, like, how we can be made available, like, on an hourly basis. And, again, it’s me, and we have a bunch of data folks on our team that do, sort of, this type of architecture work.
302 00:27:33.210 ⇒ 00:27:37.310 Uttam Kumaran: So, it’d be easy for you to just, like, Slack us, or, like.
303 00:27:37.530 ⇒ 00:27:40.299 Caleb: We can get on the line to help with architecture, like, that’s…
304 00:27:40.300 ⇒ 00:27:41.849 Uttam Kumaran: That’s a good way to leverage us.
305 00:27:42.140 ⇒ 00:27:45.379 Caleb: Yeah, also, I noticed merge… I don’t think merge.dev…
306 00:27:45.760 ⇒ 00:27:50.669 Caleb: is, like, in the… has, like, ad analytics. It’s more so, like…
307 00:27:50.980 ⇒ 00:27:52.970 Uttam Kumaran: Yeah, they may do more ERP…
308 00:27:53.050 ⇒ 00:27:54.949 Caleb: Yeah, that type of stuff, like CRM stuff.
309 00:27:54.950 ⇒ 00:27:55.780 Uttam Kumaran: Yeah.
310 00:27:56.830 ⇒ 00:28:00.829 Uttam Kumaran: But there’s a couple… there’s a lot of companies in this data broker space.
311 00:28:01.510 ⇒ 00:28:07.690 Uttam Kumaran: They all kind of just, like, are just attacks on data movement, basically.
312 00:28:08.160 ⇒ 00:28:12.370 Uttam Kumaran: So, it’s just, like, find one, and, like, find the best one for you and your clients.
313 00:28:12.720 ⇒ 00:28:13.540 Uttam Kumaran: You know.
314 00:28:19.390 ⇒ 00:28:26.190 Uttam Kumaran: Cool. Okay, well, maybe I’ll, yeah, I can shoot up an email just sort of, like, summarizing what we chatted about, and then you can kind of let me know what you think.
315 00:28:26.470 ⇒ 00:28:27.630 Caleb: Alright, sounds good.
316 00:28:27.630 ⇒ 00:28:32.400 Uttam Kumaran: What’s, like, what’s, like, sort of the timeline? Like, I haven’t… I need to call Sawyer, actually, to just hear kind of, like, what his…
317 00:28:32.560 ⇒ 00:28:37.950 Uttam Kumaran: what has… how has life been? Where are you guys at? Like, what sort of timeline on product? Like, yeah.
318 00:28:38.310 ⇒ 00:28:48.369 Caleb: Well, we’re planning on launching, or, like, we’re gonna onboard some customers this week. It’s more so, like, design partners.
319 00:28:48.370 ⇒ 00:28:48.970 Uttam Kumaran: Yay.
320 00:28:48.970 ⇒ 00:28:50.110 Caleb: And then…
321 00:28:50.870 ⇒ 00:28:59.139 Caleb: I think he’s just, like, building out the platform more, and then eventually we’ll do, like, a launch. Like, a public launch.
322 00:28:59.140 ⇒ 00:28:59.970 Uttam Kumaran: Cool.
323 00:29:00.130 ⇒ 00:29:00.900 Caleb: Yeah.
324 00:29:01.600 ⇒ 00:29:04.569 Uttam Kumaran: Yeah, okay, cool, yeah, however I can be helpful, let me know.
325 00:29:04.900 ⇒ 00:29:06.309 Caleb: Yeah, thank you so much.
326 00:29:06.600 ⇒ 00:29:08.739 Uttam Kumaran: Yeah, cool. Okay, thanks, Kale, appreciate it.
327 00:29:08.920 ⇒ 00:29:09.890 Uttam Kumaran: Okay, bye.