Meeting Title: Brainforge x Pilar Partnership Discussion Date: 2026-01-14 Meeting participants: Hoshang Mehta, Holly Condos, Uttam Kumaran
WEBVTT
1 00:00:38.070 ⇒ 00:00:41.050 Hoshang Mehta: Hey, Holly, I’m not able to hear you.
2 00:00:42.160 ⇒ 00:00:43.580 Holly Condos: Oh, hold on.
3 00:00:43.770 ⇒ 00:00:45.310 Hoshang Mehta: Yeah, can hear you now.
4 00:00:46.490 ⇒ 00:00:47.819 Holly Condos: You can hear me now?
5 00:00:47.820 ⇒ 00:00:48.789 Hoshang Mehta: Yes, I can.
6 00:00:48.790 ⇒ 00:00:51.049 Holly Condos: Okay, great. Nice to meet you.
7 00:00:51.050 ⇒ 00:00:52.759 Hoshang Mehta: Nice to meet you as well, how are you?
8 00:00:53.190 ⇒ 00:00:54.480 Holly Condos: I’m good, thanks.
9 00:00:55.440 ⇒ 00:00:56.849 Hoshang Mehta: Where are you based out of?
10 00:00:57.930 ⇒ 00:01:01.680 Holly Condos: Usually San Diego, but I’m on the East Coast this week.
11 00:01:01.680 ⇒ 00:01:02.979 Hoshang Mehta: Oh, okay, awesome.
12 00:01:02.980 ⇒ 00:01:04.190 Holly Condos: How about you?
13 00:01:04.190 ⇒ 00:01:09.650 Hoshang Mehta: Currently in India, in Bombay, but I keep shuttling between SF and Mumbai.
14 00:01:10.170 ⇒ 00:01:13.019 Holly Condos: Oh, okay. How long of a flight is that?
15 00:01:13.400 ⇒ 00:01:17.929 Hoshang Mehta: Oh, that’s anywhere around 18 to 20 hours.
16 00:01:17.930 ⇒ 00:01:19.719 Holly Condos: Oh my gosh, that’s…
17 00:01:19.720 ⇒ 00:01:25.229 Uttam Kumaran: Very long. Yeah, it’s like, you have a full day on the flight, like, breakfast, lunch, dinner, or…
18 00:01:25.230 ⇒ 00:01:28.749 Holly Condos: Dinner, breakfast, lunch, whatever.
19 00:01:28.750 ⇒ 00:01:30.269 Hoshang Mehta: Great to meet you.
20 00:01:32.500 ⇒ 00:01:33.400 Hoshang Mehta: Hey, what time, how are you?
21 00:01:33.910 ⇒ 00:01:38.580 Uttam Kumaran: Yeah, good, how are you? Thanks for taking the time. Appreciate the reach out.
22 00:01:38.970 ⇒ 00:01:46.469 Hoshang Mehta: Yeah, sorry about yesterday, I had, like, a client call that was running over, and I just decided to, move it to today. No problem.
23 00:01:46.470 ⇒ 00:01:47.880 Uttam Kumaran: All good.
24 00:01:48.220 ⇒ 00:01:48.860 Hoshang Mehta: Yeah.
25 00:01:49.220 ⇒ 00:01:51.159 Hoshang Mehta: Awesome. But.
26 00:01:51.160 ⇒ 00:02:04.940 Uttam Kumaran: Yeah, I’m more… yeah, of course, I’m more than happy to tell you about Brainforge, you know, and… but I’m really curious about the product, and yeah, just, like, interested to have a chat and see, like, you know, what we can… what we can do together. So, yeah, more than happy.
27 00:02:05.210 ⇒ 00:02:13.029 Hoshang Mehta: Yeah, yeah, definitely. So yeah, I mean, I came across Brainforge, and I kind of have been having a lot of
28 00:02:13.220 ⇒ 00:02:17.250 Hoshang Mehta: Customer conversations today, and, a lot of them…
29 00:02:17.350 ⇒ 00:02:21.600 Hoshang Mehta: you know, who are deploying AI projects, have,
30 00:02:21.660 ⇒ 00:02:40.339 Hoshang Mehta: a lot of unclear requirements in terms of how to set up their data right, right, for the AI to work. And I’m kind of seeing that there’s a need for a service component to come in and, you know, help them out, you know, with their ETL processes, or setting up their data in the correct way, so that
31 00:02:40.400 ⇒ 00:02:42.460 Hoshang Mehta: Their AI deployments work well.
32 00:02:42.470 ⇒ 00:02:52.780 Hoshang Mehta: So that’s… that’s the reason I wanted to just explore, if we can work together, but on my side, we are building Pilar, and, Pilar is…
33 00:02:52.780 ⇒ 00:03:08.549 Hoshang Mehta: basically sits in the stack between any agent tick deployments that you’re doing on NA10, Cursor, Langraph, etc, and your structured data sources. So, it’s… we sit between the data sources and the agents, and what we do is we let
34 00:03:08.550 ⇒ 00:03:25.740 Hoshang Mehta: users create, agent-facing surfaces, so that’s, like, you know, materialized views that come across different sources, write SQL queries to build those views, and then our product helps them create a MCP tool registry layer on top of it. Okay. So, you know, something like…
35 00:03:25.750 ⇒ 00:03:28.799 Hoshang Mehta: get customer tickets, could be an MCP tool that they can build.
36 00:03:28.800 ⇒ 00:03:29.380 Uttam Kumaran: I see.
37 00:03:29.630 ⇒ 00:03:31.449 Hoshang Mehta: You? Okay.
38 00:03:31.550 ⇒ 00:03:40.159 Hoshang Mehta: And that consists of, like, a parameterized SQL query, and it takes the guardrails and policies while executing at runtime. Okay.
39 00:03:40.390 ⇒ 00:03:45.629 Hoshang Mehta: So, the main problem areas that we are trying to solve for is, one is…
40 00:03:45.710 ⇒ 00:04:01.329 Hoshang Mehta: you know, if you have the most relevant data curated for the agent, it will have less probability of hallucinating and, you know, developing its own queries and, you know, doing something inconsistently. And the second is the security layer, so…
41 00:04:01.410 ⇒ 00:04:19.740 Hoshang Mehta: we have a sandboxed environment that we help users bring their data into when they’re creating agent surfaces. So whenever, like, you know, if there’s an attack happening, it doesn’t know where the data comes from, it has no way of entering the source tables in your Databricks or Snowflake.
42 00:04:19.740 ⇒ 00:04:25.439 Hoshang Mehta: It roams around in the sandbox, and it could… the blast radius of an attack is very minimal.
43 00:04:25.440 ⇒ 00:04:36.089 Hoshang Mehta: Okay. So, yeah, along with all of that, there’s detailed around, you know, policies and guidance that we enforce as well at runtime. But in a nutshell, that’s basically where we are going.
44 00:04:36.530 ⇒ 00:04:49.600 Uttam Kumaran: Okay, cool, I mean, like I said, actually, a lot of work that we do, so I’m actually very curious about, like, maybe we can even bring some of our customers to use your product. We do a lot of what you described, you know, in some situations, very manually these days.
45 00:04:49.600 ⇒ 00:04:59.050 Uttam Kumaran: A little bit of background on us, like, we’re a data and AI consultancy, so my background is in data engineering, work for a number of firms and startups leading data teams.
46 00:04:59.050 ⇒ 00:05:13.660 Uttam Kumaran: And then leading product at a data startup. So data is really, like, our bread and butter. It just so happens that I also built a lot of the company using AI over the last few years, and then naturally, we were like, hey, we can go help other companies also build agents.
47 00:05:13.670 ⇒ 00:05:23.950 Uttam Kumaran: MCP, chat, most of our… our support is typically on the internal side, so we’re building internal chatbots. But again, it’s very similar in that
48 00:05:23.950 ⇒ 00:05:36.820 Uttam Kumaran: In some situations, we’re using MCP, in some situations, we’re landing data. We’ve built, like, Texas SQL sort of interfaces before as well. But yeah, a lot of our clients are facing the same thing, and so for us, yes, one is, like, we…
49 00:05:36.820 ⇒ 00:05:46.679 Uttam Kumaran: create very pretty clean tables for agents to query with, like, great metadata. That way, it’s, like, it usually does a great job, versus if you just give it, like.
50 00:05:46.780 ⇒ 00:06:01.060 Uttam Kumaran: the… basically, you just land data, you give it, like, the… whatever it’s raw there, it’s gonna mess up. Unfortunately, though, clients will blame your product, they won’t blame the fact that they didn’t… they didn’t set up the context in the right way, and so that’s really…
51 00:06:01.060 ⇒ 00:06:05.400 Uttam Kumaran: why I think we do a good job is that our background is fully in
52 00:06:05.400 ⇒ 00:06:26.870 Uttam Kumaran: you know, preparing data sets, so this is, like, landing data into any sort of data warehouse or database, so we do a lot of work with, like, Snowflake, for example. This is modeling it, so using dbt to, like, model a lot of, like, you know, core reporting marts, and then making sure that is made available to agents, via, via MCP or via, like, Texas SQL.
53 00:06:26.870 ⇒ 00:06:43.970 Uttam Kumaran: you know, sort of, like, methods. And then, additionally, it’s really, like, the prompt engineering is actually, like, kind of the easiest part, and picking the LLM is the easiest part. Then it’s more on the integration side. So, is the data going to… is the data going to Slack, so we need, like, rich Slack interfaces.
54 00:06:43.970 ⇒ 00:07:02.269 Uttam Kumaran: Is it in a chat? So we use, like, CopilotKit, or we use, like, Mastra, or V0, like, we think about, like, what is the output environment, but those are sort of, like, where… both of those are, like, data problems, you know, much less of a AI, like, LLM problem. But in fact, it’s, like, it really affects, like, the user experience of any, like.
55 00:07:02.420 ⇒ 00:07:04.549 Uttam Kumaran: sort of interface, AI interface, you know?
56 00:07:04.780 ⇒ 00:07:18.849 Hoshang Mehta: Correct, correct. No, that makes a lot of sense. So, just so that I understand it correctly, you do the data infra setup work, as well as you build the agents for them, like, on an NA10 or a Langraph, or is it more on the data side?
57 00:07:18.850 ⇒ 00:07:31.839 Uttam Kumaran: No, it’s both. We do both. So we… so we have about 25 people, like, kind of scattered, a lot of folks here in the States and, like, in a couple different places. But we… so we have a few, we have, like, 4 or 5 full-stack folks.
58 00:07:31.840 ⇒ 00:07:40.120 Uttam Kumaran: As well as, like, a whole team of data folks. So, you know, the kind of the data org is really split up from data engineering to sort of more, like, analytics engineering, modeling.
59 00:07:40.120 ⇒ 00:07:51.299 Uttam Kumaran: and we also do, like, strategy, product analytics, so, like, amplitude mixed panel type stuff. And then, on the AI side, yeah, it’s really, like, sort of full-stack development, plus all of the observability around
60 00:07:51.300 ⇒ 00:08:09.010 Uttam Kumaran: all the observability and, you know, customization on the AI side. So, for example, we have some clients that’s more document-heavy, so we’re using different, like, you know, document retrieval, you know, services to produce that. We do a lot of NADEN work, although most of these days, we’re moving a lot to Mostra.
61 00:08:09.230 ⇒ 00:08:17.080 Uttam Kumaran: And we’re just finding it, like, a lot more scalable, a lot easier to do observability, a lot easier to catch bugs, although
62 00:08:17.080 ⇒ 00:08:32.350 Uttam Kumaran: our company, we’ve done a lot of NEDEN work, so we’re kind of not afraid of that, but typically, as these systems get big, it’s just, like, natural to move into more of, like, a traditional environment. So yeah, we do both, and really, like, I think our advantage is that we do both, like.
63 00:08:32.350 ⇒ 00:08:50.909 Uttam Kumaran: We have a really deep data experience, and we, like, we also use a lot of these AI tools internally, and we build a lot for ourselves, so it made it really easy to go to a client and being like, hey, you’re probably trying to prototype some of these, and you’re getting limited because you don’t have a great input data set, or you haven’t thought about
64 00:08:50.910 ⇒ 00:09:10.910 Uttam Kumaran: the knowledge engineering or context engineering. And then second is, like, okay, yes, you can just enable some quick chat environment, but you need to, for example, do, like, scheduled reports, you need to get this into another application, or you need to get this into Slack, right? Those are, like, integrations that have to get built where structured data is getting sent
65 00:09:10.910 ⇒ 00:09:17.250 Uttam Kumaran: somewhere, in a reliable manner. So, anything around data movement, orchestration, modeling is, like.
66 00:09:17.420 ⇒ 00:09:32.490 Uttam Kumaran: That’s, like, really, like, what we’re really, really great at. And then the AI piece just came up because that’s the new medium to interact with a lot of information. So even a lot of our clients, when we come in and we do traditional reporting, they always are now asking, like, hey, how can I start to access some of the insights?
67 00:09:32.580 ⇒ 00:09:43.099 Uttam Kumaran: and data via AI. So, sometimes we recommend there’s some really great BI tools out there that have out-of-the-box AI. Other times, we’re also just building simple interfaces to
68 00:09:43.170 ⇒ 00:09:51.230 Uttam Kumaran: you know, basically grab insights or grab stuff from the summary tables that we build. And the summary tables have really rich context, right? Like.
69 00:09:51.230 ⇒ 00:10:04.380 Uttam Kumaran: If the AI is able to actually see the SQL code underneath, it’s able to infer, like, what the transformations are going to. Of course, we have, like, document… column level, table level, documentation, source level.
70 00:10:04.380 ⇒ 00:10:08.059 Uttam Kumaran: documentation. And so, it tends to actually…
71 00:10:08.210 ⇒ 00:10:23.749 Uttam Kumaran: it tends to be working, like, super, super well for the majority of questions, but you do have to provide it with that context. Like, just throwing it over a raw, like, database, it’s gonna have no clue, you know? Or it’s gonna… it’s basically gonna assume, and it’s gonna get it wrong, and it’s like a…
72 00:10:23.980 ⇒ 00:10:35.879 Uttam Kumaran: We really try to come in at the time for clients when they’ve tried the solution, maybe they’re not getting further enough, or when they’re just, like, embarking on this, because then we can sort of lead them down, you know, a really, really great path.
73 00:10:35.880 ⇒ 00:10:47.349 Hoshang Mehta: Yeah, yeah, exactly. No, that makes sense. I think that’s quite unique about the service that you offer, is more data first, and then AI is just a medium to kind of interact with all the.
74 00:10:47.350 ⇒ 00:10:50.399 Uttam Kumaran: Because we’re not… we’re not doing any, like, crazy advanced…
75 00:10:50.410 ⇒ 00:11:10.369 Uttam Kumaran: AI capabilities, and again, we’re a service company, so I’m not here to build software. I don’t want to build… I want to use tools like yours. So, for us, it’s unfortunate that we’ve had to build a lot of this custom, but now there’s a lot more tools coming to market that allow us to just move faster, and move faster towards an accurate response.
76 00:11:10.850 ⇒ 00:11:11.680 Uttam Kumaran: customers.
77 00:11:11.680 ⇒ 00:11:14.450 Hoshang Mehta: But yes, exactly, like, I felt like…
78 00:11:14.450 ⇒ 00:11:18.260 Uttam Kumaran: Naturally, like, we realized that actually most of this problem is, like, an input
79 00:11:18.510 ⇒ 00:11:29.460 Uttam Kumaran: problem, and then, like, an output problem. In the middle is, like, we… we rely on, like, the LLMs, like, different orchestration tools, different frameworks to sort of handle… handle that.
80 00:11:29.710 ⇒ 00:11:47.350 Hoshang Mehta: Correct. Got it, got it. Now, that makes a lot of sense. I think we are aligned there. I also wanted to assess if, like, we both have the same customer base that we are trying to address, in terms of, like, you know, what’s your ICP in terms of who do you, like, best fit for?
81 00:11:47.650 ⇒ 00:11:55.410 Uttam Kumaran: Yeah, so, I mean, typically, we’re working with… we’re working with companies that are anywhere from $20 million in revenue upwards.
82 00:11:55.420 ⇒ 00:11:57.660 Uttam Kumaran: So we have a… we have a bunch of people…
83 00:11:57.660 ⇒ 00:12:19.250 Uttam Kumaran: sort of, like, in the 20 to 100 million, and then we have a few brands that we’re working with that are in the couple hundred million. My background is in startups, we’ve worked with a lot of startups. As you probably know, it’s just not, like, the most sticky, best customer. I assume you guys are also trying to look to kind of gain enterprise and move higher. So that’s really, like, where we land. So, one is, like.
84 00:12:19.250 ⇒ 00:12:39.110 Uttam Kumaran: we’re coming into companies with real need, they want to grow, and they have budget to do so, but maybe they lack the internal talent, maybe they lack the internal pace, right? Maybe they do have some team, but they just don’t have the pace to kind of move forward. Or there’s, like, some really timely thing they have to hit, where they’re just like, we need, like, a Navy SEAL kind of team.
85 00:12:39.180 ⇒ 00:12:58.309 Uttam Kumaran: And so, we have done work with… with some startups in the past. Typically, they have to be, like, well-funded and… and, like, usually… because otherwise, we… when we built the business on startups, it was just a tough customer, you know? And I worked… I was a tough customer as a… in startups as well, so that’s really, like, where…
86 00:12:58.310 ⇒ 00:13:02.519 Uttam Kumaran: Where we’re focused on, and we’re starting to move further and further up.
87 00:13:02.610 ⇒ 00:13:03.860 Uttam Kumaran: You know, the chain.
88 00:13:04.050 ⇒ 00:13:09.240 Hoshang Mehta: Got it. And in terms of industry, is it… is it like B2B SaaS, B2C SaaS, or is there…
89 00:13:09.370 ⇒ 00:13:10.520 Hoshang Mehta: agnostic.
90 00:13:10.850 ⇒ 00:13:24.630 Uttam Kumaran: Yeah, so we, and I’ll even just pull up, like, a couple slides so you can kind of see this, but, we… we work with, typically, a lot of our background, me and my business partners, in B2B, SaaS, and e-commerce.
91 00:13:24.670 ⇒ 00:13:35.900 Uttam Kumaran: So we worked in a lot in B2B SaaS, now a lot in B2C, in e-com, so e-com, omnichannel, retail, CPG, and then we’re also starting to do a lot in health.
92 00:13:35.940 ⇒ 00:13:41.979 Uttam Kumaran: I would say, like, we’re trying to work… we’re trying to, of course, move more enterprise, so…
93 00:13:42.060 ⇒ 00:14:00.969 Uttam Kumaran: work with larger clients, but also work with larger clients where their need is really, really high, where they may not be able to get the talent, or if they… even if they have the talent, they may not be privy to, like, what’s the latest in AI. So that’s kind of how we’re heading, although we have a… we’ve worked with quite a bit of clients across a variety of industries, so yeah.
94 00:14:00.970 ⇒ 00:14:06.970 Hoshang Mehta: Got it. No, that makes sense. I think quite aligned, there. We… our customer base is also…
95 00:14:07.220 ⇒ 00:14:17.529 Hoshang Mehta: more around, like, if you know, in a very raw sense, industries that have a lot of structured data, and all of this data is either custom or PII, private.
96 00:14:17.530 ⇒ 00:14:18.140 Uttam Kumaran: Okay.
97 00:14:18.490 ⇒ 00:14:25.620 Hoshang Mehta: And they’re building AI to kind of, you know, make their operations easy, whether it’s internal chatbots, analytics.
98 00:14:25.790 ⇒ 00:14:38.000 Hoshang Mehta: And a few of our customers also are building a customer-facing chatbot on their tools, for, you know, their customers to kind of make sense out of the insights that their tool provides.
99 00:14:38.000 ⇒ 00:14:51.210 Uttam Kumaran: Tell me about, like, the timing in which you guys enter. Like, do you find your customers have already tried something, or do you find that, like, you’re the pathway to actually, like, realizing the value?
100 00:14:51.880 ⇒ 00:15:06.820 Hoshang Mehta: No, so, good question. So, we’ve had some serious, you know, talking about a more mature customer that, you know, knows exactly what they’re looking for at the budget that they have. They have already tried,
101 00:15:07.370 ⇒ 00:15:27.349 Hoshang Mehta: something or the other in terms of, you know, making the project move, whether it’s, like, building their own set of APIs to their data sources and cleaning up the data. So, they’ve already kind of started preparing their data in some sense for the AI workload, is where we kind of come in and be like, okay, you know, you can move fast from your
102 00:15:27.350 ⇒ 00:15:37.729 Hoshang Mehta: But at the same time, your projects, like in e-commerce, for example, or fintech as well, a lot of their projects get, held up in, security.
103 00:15:37.730 ⇒ 00:15:39.860 Uttam Kumaran: side of things, like, hey… Okay.
104 00:15:40.120 ⇒ 00:15:41.829 Hoshang Mehta: what if this LLM is gonna…
105 00:15:41.830 ⇒ 00:15:42.390 Uttam Kumaran: Yes.
106 00:15:42.550 ⇒ 00:16:01.830 Hoshang Mehta: a customer data point. And then they kind of go back and forth there and try to resolve it with, like, architecture, solutions, etc. So that’s why we come in. So there’s some kind of movement in terms of exactly what they want to solve for with AI. They’ve started to prepare their data, they understand the data side of it is important.
107 00:16:01.830 ⇒ 00:16:05.870 Hoshang Mehta: Is when we come in and we make this process seamless for them.
108 00:16:06.370 ⇒ 00:16:17.320 Uttam Kumaran: Okay, okay. And do you feel like the… like, are you typically selling into, like, the technical organization, or are you selling into the business side of the organization, or, like, what has it been so far?
109 00:16:17.610 ⇒ 00:16:35.729 Hoshang Mehta: It’s been a mix, man, like, a few times… most other times, it’s been around the technical, you know, bottoms-up movement. More along, like, you know, head of product, or CTO kind of understanding what we are doing and relaying back to their team.
110 00:16:35.730 ⇒ 00:16:36.190 Uttam Kumaran: Okay.
111 00:16:36.190 ⇒ 00:16:40.409 Hoshang Mehta: Few of our customers have also started coming from
112 00:16:40.600 ⇒ 00:16:59.719 Hoshang Mehta: the head of operations or, like, business side of, things, because they understand the business impact of what they imagine the AI to solve for, and then they’re looking for, like, tools. But it’s a mix, like, but 60% is coming from the tech organizations within these companies.
113 00:17:00.040 ⇒ 00:17:07.260 Uttam Kumaran: Okay, okay, okay, great. Yeah, one… maybe one, like, slide I’ll just share… I’ll just share with you is, like, this is sort of, like.
114 00:17:07.280 ⇒ 00:17:26.980 Uttam Kumaran: one of our AI, like, capabilities decks that I’ll send to you, but, like, this is kind of, like, how we describe, you know, our solution, typically. It’s like, we have a bunch of data sources, which is, like, either structured data, it’s either these, like, operational sort of platforms, it can be unstructured data that sits in a lake.
115 00:17:26.980 ⇒ 00:17:29.140 Hoshang Mehta: We… we push it into the.
116 00:17:29.140 ⇒ 00:17:39.100 Uttam Kumaran: you know, a place to actually execute things. And then, of course, where we lean on is that we spend a lot of time on, like, observability. And so, like, we basically come in and we’re like, look.
117 00:17:39.510 ⇒ 00:17:45.570 Uttam Kumaran: at this point, like, so when I was selling AI services, like, Jan of last year.
118 00:17:45.570 ⇒ 00:17:47.910 Hoshang Mehta: Nobody was even, like… people were just like.
119 00:17:47.910 ⇒ 00:17:53.100 Uttam Kumaran: don’t even know where to go, and so it was, like, really rough. Because we had been using, like.
120 00:17:53.110 ⇒ 00:18:10.750 Uttam Kumaran: actually, none of… none of this has changed in our business, but, like, we were really early, you know? And so when… it was… it was kind of early for us to even talk about how, like, you can’t go past the prototyping stage without, like, thinking about a solution. But now, this is really hitting, because most…
121 00:18:10.750 ⇒ 00:18:17.869 Uttam Kumaran: Like, companies that are serious about this have already decided to put budget towards it, have already built, like, the committee, and maybe tried something.
122 00:18:17.870 ⇒ 00:18:31.040 Uttam Kumaran: But naturally, they’ve hit a roadblock, probably for one of these reasons, is they… they tried something, they bought a vendor, anecdotally, the demo was really flashy and nice, but then they don’t know if it’s getting used, they don’t know how to improve it.
123 00:18:31.090 ⇒ 00:18:46.489 Uttam Kumaran: they’re probably… it’s like, it’s actually not working like they thought, because their input data sucks. The vendor, you know, is kind of the reason why we’re talking, is there’s not much professional services available, typically, to do the curation of the input data, which affects
124 00:18:46.490 ⇒ 00:18:58.550 Uttam Kumaran: your product, right, and it fucks the adoption of your product, they don’t have the capability of doing that. And then that’s kind of where we come in, where we’re like, look, we want to make sure that, like, you have really, like, great context.
125 00:18:58.550 ⇒ 00:19:11.510 Uttam Kumaran: we do evals, and we make sure that we’re reporting on the adoption, right? So in your situation, like, we would report on the… how many MCPs exist, which are getting called the most, right? And so that’s what, like.
126 00:19:11.520 ⇒ 00:19:16.440 Uttam Kumaran: Those are the success metrics of this, like, entire initiative for the client.
127 00:19:16.470 ⇒ 00:19:19.919 Uttam Kumaran: And so, like, we kind of walked them through, basically, like.
128 00:19:20.100 ⇒ 00:19:30.409 Uttam Kumaran: basically, you’re either… you don’t even know where to start, or you’re making… you made investments. We sort of talk about how we show metrics around usage, leaderboards, so it’s like…
129 00:19:30.410 ⇒ 00:19:41.859 Uttam Kumaran: broader, broader, you know, adoption of AI within a solution. We talk about evals, we talk about, like, embedding, you know, things into your workflow,
130 00:19:41.910 ⇒ 00:19:46.299 Uttam Kumaran: And then we sort of talk about, like, okay, what are some of the tools that we’re familiar with, governance.
131 00:19:46.300 ⇒ 00:19:48.029 Hoshang Mehta: But then this is really where, like.
132 00:19:48.030 ⇒ 00:20:04.339 Uttam Kumaran: we talk about just, here’s the overall, like, you know, architecture. But it seems pretty in line, you know, I’m happy to send you this deck. This is, you know, and this is basically what we share when we’re talking to clients. And then additionally, again, with all of these.
133 00:20:04.340 ⇒ 00:20:15.559 Uttam Kumaran: sort of folks in this value chain we’ve established really great partnerships with. Which means, like, we’re quick… we can get customers really great discounts for, like, Snowflake, or for Mother Duck.
134 00:20:15.600 ⇒ 00:20:25.900 Uttam Kumaran: You know, for different data warehousing, dbt, ETL tools, because it’s not a small undertaking to, like, land and model your data effectively, but it is, like.
135 00:20:26.000 ⇒ 00:20:36.479 Uttam Kumaran: completely, like, the bottleneck, typically. And you want more of your… the team to be focused on your software and understanding the ergonomics there, and to just, like, nail this, you know?
136 00:20:36.750 ⇒ 00:20:42.119 Uttam Kumaran: And so, like, I think we’re… I think we’re, you know, fairly aligned, so…
137 00:20:42.120 ⇒ 00:20:42.840 Hoshang Mehta: Yeah.
138 00:20:43.010 ⇒ 00:20:56.980 Hoshang Mehta: No, makes sense. This is quite, cool. Cool, let me also show you a peek of what we are doing, and maybe, you know, we can come up with some ideas there as well, in terms of where you see the use.
139 00:20:57.250 ⇒ 00:20:59.880 Hoshang Mehta: System…
140 00:21:03.320 ⇒ 00:21:04.860 Hoshang Mehta: Can you see my screen?
141 00:21:05.130 ⇒ 00:21:05.780 Uttam Kumaran: Yes.
142 00:21:07.360 ⇒ 00:21:22.790 Hoshang Mehta: All right, so first thing is that we handle these integrations, right? All the structured sources. So, you have all your databases, data warehouses that you can connect to one or more of them. We also…
143 00:21:22.790 ⇒ 00:21:35.100 Hoshang Mehta: offer ETL underneath, so you can connect to any of these business connectors as well. And what we do is we centralize all your data in one place, under the product. So let’s say that you’ve connected to
144 00:21:35.160 ⇒ 00:21:47.889 Hoshang Mehta: Google Analytics, PostHoc, BigQuery, and you want to now kind of build an agent that looks at data across these different sources, to make sense out of the work that it’s assigned to do.
145 00:21:48.130 ⇒ 00:21:49.700 Hoshang Mehta: So, for example.
146 00:21:50.250 ⇒ 00:22:01.680 Hoshang Mehta: this is where the data curation part happens. So on the left, you see, like, I have my connected sources, and here’s where I write my SQL and generate my views, so I can…
147 00:22:01.680 ⇒ 00:22:12.819 Hoshang Mehta: Say, I write one query to HubSpot, one to Zendesk, the other to Amplitude, and now I’ve created, like, a combined… I’ve joined across these sources to create a view.
148 00:22:12.830 ⇒ 00:22:16.390 Hoshang Mehta: That I want, specifically, my LLM to interact with.
149 00:22:16.420 ⇒ 00:22:26.590 Hoshang Mehta: So this is the agent surface that you create via SQL queries. And now this is sandboxed, so we spin up a sandbox environment so the LLM
150 00:22:26.590 ⇒ 00:22:37.350 Hoshang Mehta: hitting this sandbox only knows that these are the fields, and this is the context I’m supposed to work with. It doesn’t go roaming around your database warehouses and, you know, make sense out of.
151 00:22:37.350 ⇒ 00:22:40.170 Uttam Kumaran: Or, like, using a bunch of tools, like, it doesn’t need to, okay.
152 00:22:40.170 ⇒ 00:22:55.389 Hoshang Mehta: Correct, correct. And then we have an AI MCP tool builder, which you can do manually as well, but what it looks like is this. So, these are all the MCP tools that I can chat with RAI, to create. So, let’s say I want to…
153 00:22:55.390 ⇒ 00:23:06.290 Hoshang Mehta: analyze customer health patterns. So, it spins up an entire registry with the function name, description, a parameterized SQL query, and you can also set policies here.
154 00:23:06.470 ⇒ 00:23:13.040 Hoshang Mehta: Which you can test and, you know, do all sorts of things. So, for example, here.
155 00:23:13.150 ⇒ 00:23:19.540 Hoshang Mehta: I can do a test run, and I can… I have a policy to not let customer A access customer B’s information.
156 00:23:19.540 ⇒ 00:23:23.370 Uttam Kumaran: Sure. So it inputs the filter based on the input user. Okay.
157 00:23:23.370 ⇒ 00:23:25.699 Hoshang Mehta: So if I’m impersonating someone else.
158 00:23:25.700 ⇒ 00:23:27.570 Uttam Kumaran: their attributes or something, yeah.
159 00:23:27.570 ⇒ 00:23:30.520 Hoshang Mehta: Yeah, so it’ll run into a policy error here.
160 00:23:30.770 ⇒ 00:23:36.100 Hoshang Mehta: And then we have a preview agent, which kind of is a sandboxed,
161 00:23:36.140 ⇒ 00:23:53.590 Hoshang Mehta: you know, test version of how you can interact with, like, impersonating how an actual user would use the product, whether it’s in a chatbot. You prompt, and it’ll… you’ll see that it calls the right tools and get you the right information, which you can deploy to prod from here in your pipeline. So.
162 00:23:53.590 ⇒ 00:23:54.210 Uttam Kumaran: Okay.
163 00:23:54.210 ⇒ 00:24:04.010 Hoshang Mehta: And it’s agnostic to any agent builder that you use, so you can select all of these tools that you’ve built, and with a…
164 00:24:04.510 ⇒ 00:24:16.120 Hoshang Mehta: with a token and an MCP endpoint URL, you can push it to one or more agents, like, whether you want agents from any 10 to interact with this dataset, plus Langra.
165 00:24:16.120 ⇒ 00:24:16.609 Uttam Kumaran: I see.
166 00:24:16.610 ⇒ 00:24:24.600 Hoshang Mehta: Let’s make. So you can do all sorts of things here, and, you also get, an evals in terms of
167 00:24:24.690 ⇒ 00:24:40.620 Hoshang Mehta: How many calls were made, if something… a policy was violated, you can see the query that it ran. And, you know, the evils that you were talking about is basically, like, you can now see whether you want to increase or decrease the scope of your view.
168 00:24:40.620 ⇒ 00:24:52.239 Hoshang Mehta: or add one more MCP tool to fetch the right information. You can do it inside this one control plane, and it goes and reflects in all the agents that you’ve already published these tools into.
169 00:24:52.490 ⇒ 00:25:05.269 Uttam Kumaran: I see, okay, okay. So, ultimately, like, if I was to save back, basically, instead of just giving, like, an agent with, like, Texas SQL tool calling on this, you’re almost giving it, like.
170 00:25:05.630 ⇒ 00:25:13.429 Uttam Kumaran: you’re almost giving it, like, access to, like, a summary or a certain join of certain things to kind of limit the context. Yeah. And then…
171 00:25:13.600 ⇒ 00:25:18.779 Uttam Kumaran: It’s creating the… is it creating… so it’s creating a materialized view, and then…
172 00:25:19.010 ⇒ 00:25:23.200 Uttam Kumaran: Bringing in parts of that into context, or, like, how does that, like, handoff work?
173 00:25:24.060 ⇒ 00:25:32.380 Hoshang Mehta: So via MCP Tools is basically where… how we… how we, how the users let the LLM access this data.
174 00:25:32.380 ⇒ 00:25:33.120 Uttam Kumaran: Okay.
175 00:25:33.120 ⇒ 00:25:41.370 Hoshang Mehta: So, let’s say that… and all of this is coming from a point of view that a lot of our customers start in a very deterministic manner.
176 00:25:41.470 ⇒ 00:25:53.129 Hoshang Mehta: And then they kind of go more autonomous… on the autonomous side. So, they kind of create interaction mediums in terms of, okay, this tool is used for…
177 00:25:53.130 ⇒ 00:26:09.040 Hoshang Mehta: assessing what’s the revenue impact, and it needs an account ID. So it’ll go and fetch that account ID from the user, and then pass it to the query, check the policy at runtime before the LLM hits the data and queries it, and then gets the result back to the agent.
178 00:26:09.360 ⇒ 00:26:09.900 Uttam Kumaran: Okay.
179 00:26:10.010 ⇒ 00:26:10.750 Uttam Kumaran: Okay.
180 00:26:11.600 ⇒ 00:26:12.190 Hoshang Mehta: Yeah.
181 00:26:12.610 ⇒ 00:26:18.189 Uttam Kumaran: And then, is it just… SQL, or are you gonna… are you working with, like, unstructured…
182 00:26:18.460 ⇒ 00:26:33.610 Hoshang Mehta: Yeah, unstructured is coming up. At the moment, we are focusing more on structured sources, but unstructured on the architecture side, we are working through some things on, like, vectorizing, converting, you know, unstructured to structured.
183 00:26:34.020 ⇒ 00:26:41.049 Hoshang Mehta: I think these things work on it. But at the moment, we are focused on, like, just creating a wedge in the structured data to AI space.
184 00:26:41.460 ⇒ 00:26:47.040 Uttam Kumaran: Okay. Yeah, and then where, like, where are you guys seeing, like, use case wins?
185 00:26:47.610 ⇒ 00:26:55.160 Hoshang Mehta: So, a lot of these, customers that we are working with do come from e-commerce, FinTech.
186 00:26:55.310 ⇒ 00:27:01.170 Hoshang Mehta: companies. So, they do have a lot of structured data, and all of it is PII, personal data.
187 00:27:01.170 ⇒ 00:27:02.289 Uttam Kumaran: Yeah, yeah, yeah.
188 00:27:02.290 ⇒ 00:27:09.989 Hoshang Mehta: So, and they are building AI to, you know, whether it’s… most of them are internal, operations, analytics kind of use cases.
189 00:27:09.990 ⇒ 00:27:14.689 Uttam Kumaran: Yeah, like, ID lookups, like, tell me what this order is, tell me about this customer, yeah.
190 00:27:14.690 ⇒ 00:27:19.489 Hoshang Mehta: So, like, even support, bots, so…
191 00:27:19.490 ⇒ 00:27:24.219 Uttam Kumaran: And they’re building this in order… cause… because there are a wealth of other…
192 00:27:24.340 ⇒ 00:27:29.640 Uttam Kumaran: vendors out there that do, like, customer service agents, right? Like, are they building this on their own? Because
193 00:27:29.810 ⇒ 00:27:32.180 Uttam Kumaran: It’s like a governance thing, or it’s like…
194 00:27:32.640 ⇒ 00:27:36.440 Uttam Kumaran: this is cheaper, like, what do you think… what do you think it is? Like, what’s the narrative?
195 00:27:36.440 ⇒ 00:27:46.149 Hoshang Mehta: Yeah, so smaller companies do go for, you know, the customer service agent platforms that already exist.
196 00:27:46.270 ⇒ 00:27:52.219 Hoshang Mehta: But as we go upmarket, there is a lot of customization, privacy, guardrails.
197 00:27:52.220 ⇒ 00:27:53.590 Uttam Kumaran: Yeah, yeah, okay, great.
198 00:27:54.030 ⇒ 00:28:13.169 Hoshang Mehta: And like you said, like, these tools are, not, they don’t have the support to build all this for them. So that’s exactly what I’m seeing as well, which is why you’re seeing a more pull from, someone who wants to customize, but they also have, like, standard agent building toolkits from, like, say, Langraph or Agnova.
199 00:28:13.170 ⇒ 00:28:13.790 Uttam Kumaran: Yes.
200 00:28:13.790 ⇒ 00:28:15.919 Hoshang Mehta: that they’re building on. Yeah.
201 00:28:16.220 ⇒ 00:28:34.439 Hoshang Mehta: So… and also, a recent realization is that a lot of them, as we go upmarket, are using different agents. Like, a marketing team is using, say, an NA10, but the product is using Langraph, and someone else is using something else.
202 00:28:34.750 ⇒ 00:28:47.700 Uttam Kumaran: You know what’s gonna happen? As, like, Claude for work or Cursor gets more popular, like, we… a lot of our business teams use Cursor, because they’re writing a lot, and they’re still… so they’re doing MCP calls to Notion, to Linear.
203 00:28:47.910 ⇒ 00:28:53.030 Uttam Kumaran: And then so we’re… and then so we’re… right now, we’re writing our own, like, Google Drive…
204 00:28:53.060 ⇒ 00:29:08.989 Uttam Kumaran: one, because there’s not, like, an official one, right? But this would be a great example of, like, we should just… we could use this for that, and there’s a lot of other structured data use cases. For us, again, right now, there isn’t a solution for this that we recommend, because typically we’re doing stuff in Snowflake.
205 00:29:09.080 ⇒ 00:29:18.750 Uttam Kumaran: Snowflake’s, like, inbuilt AI is, like, not good, and so what we’re tasked… what we basically do is we create, like, the summary table of, like.
206 00:29:18.860 ⇒ 00:29:29.229 Uttam Kumaran: customers and revenue, and then we just build Text2SQL directly to that, instead of… but then we… it’s usually, like, just narrow, it’s, like, one or two things. We’re not…
207 00:29:29.320 ⇒ 00:29:40.359 Uttam Kumaran: I think this is a great way to support, like, tons of tool use cases, and, like, create a layer there to manage, versus just giving an agent access to, like.
208 00:29:40.910 ⇒ 00:29:45.559 Uttam Kumaran: a bunch of tables, right? And not really being able to steer it. Is that, like, a good way to, like…
209 00:29:45.760 ⇒ 00:29:51.350 Hoshang Mehta: Yeah, exactly. So, the… I mean, when you do NLP to SQL,
210 00:29:51.350 ⇒ 00:30:10.030 Hoshang Mehta: You might see certain cases where, like, it’s giving you one answer today, and then it’ll give you another answer tomorrow. But, I mean, it’s not… you don’t have any control over what it’s querying, whether it’s supposed to query this, whether it’s supposed to check these things, and these are the fields that are healthy.
211 00:30:10.030 ⇒ 00:30:28.599 Hoshang Mehta: All of those things. So, all of those things can be set via the MCP tool layer that are built on our views, and you can be adding n number of them. So, whenever the agent is deciding where my information is, it just has to decide, okay, this is the tool I have to call, and then it queries the exact situation.
212 00:30:28.600 ⇒ 00:30:29.030 Uttam Kumaran: Yeah.
213 00:30:29.030 ⇒ 00:30:29.380 Hoshang Mehta: Yeah.
214 00:30:29.380 ⇒ 00:30:32.129 Uttam Kumaran: Great, great. Yeah, it’s much more narrow.
215 00:30:32.130 ⇒ 00:30:32.750 Hoshang Mehta: Yeah.
216 00:30:32.750 ⇒ 00:30:37.560 Uttam Kumaran: Versus, like, doing exploratory, you know, which is actually most of the use cases. Okay.
217 00:30:37.610 ⇒ 00:30:53.969 Uttam Kumaran: Yeah, this is great. I mean, I love the product. We even have some folks that, like, I would love to even have our team try it for, you know? But, yeah, tell me, like, what’s good next steps? I mean, we’re… we’re in Slack. I don’t know if you guys are also in Slack. We can totally start, like, a shared workspace.
218 00:30:54.230 ⇒ 00:31:05.220 Uttam Kumaran: And then, yeah, happy to… if you want to let me know if there’s any customers that you think, on your side that need help urgently, like, we can kind of tell you, like, what our typical…
219 00:31:05.270 ⇒ 00:31:14.890 Uttam Kumaran: method of attack is, and sort of even send you case studies and some stuff about what we’ve done, and… and maybe, like, go from there. But would certainly be open to helping, and…
220 00:31:15.080 ⇒ 00:31:27.229 Uttam Kumaran: I would say, looking at the product, our whole team would… is totally capable of… of leveraging and building on top of that, but really, again, we’re… we’re probably doing the before half, is just, like, making sure all of that is set up so that
221 00:31:27.350 ⇒ 00:31:33.820 Uttam Kumaran: when you go to create the MCPs, it’s really easy to run, like, simple selects on clean data marks.
222 00:31:34.070 ⇒ 00:31:39.689 Hoshang Mehta: Correct. So that’s exactly where I’m seeing this collaboration happen, because when I go to a customer
223 00:31:39.810 ⇒ 00:31:46.730 Hoshang Mehta: Every now and then, there is someone who’s, you know, gonna be like, hey, but we need to kind of set up our data right for.
224 00:31:46.730 ⇒ 00:31:47.090 Uttam Kumaran: Yeah.
225 00:31:47.090 ⇒ 00:31:49.890 Hoshang Mehta: bringing it to Killer, and then have the AI work on it.
226 00:31:49.930 ⇒ 00:32:09.800 Hoshang Mehta: And that’s when I want to kind of say that, hey, I have a partner that can help you speed this up, and here’s how they’ll work. So, it gets their, conviction also go higher, and, we can work together as well on that front, is the reason why I’m thinking of, like, collaborating with you guys.
227 00:32:09.800 ⇒ 00:32:15.119 Uttam Kumaran: Yeah, this is very similar to, like, when we work with several other vendors, like, for example, even, like.
228 00:32:15.260 ⇒ 00:32:27.920 Uttam Kumaran: we work with ETL vendors all the way to BI tools, and in both situations, like, they just do one part of the value chain, right? And they go to clients, and they’re like, hey, we can land all your data, and then they’re like, yeah, but can you help us get to the clean tables? They’re like.
229 00:32:27.920 ⇒ 00:32:39.249 Uttam Kumaran: That’s not, like, our job. We’re just, like, the broker. Same with the BI tools, right? Like, Omni is one of our partners. They’ll go in, they’ll be like, you guys don’t have a team to manage this, like, you don’t have clean tables, so we’re…
230 00:32:39.390 ⇒ 00:32:52.110 Uttam Kumaran: yes, you could buy… we could sell you our tool, but you’re just gonna put it on, like, raw data, and it’s gonna be, like, kind of useless. And so that’s exactly where they bring us in, and I feel like it’s one of the ways that our business has grown a lot, is because we just work
231 00:32:52.160 ⇒ 00:33:01.530 Uttam Kumaran: Super, super well with our vendor partners, and basically, most of our deals these days are being done in association with one or many vendor partners.
232 00:33:01.530 ⇒ 00:33:04.290 Hoshang Mehta: Because it’s just, like, so much nicer to give a…
233 00:33:04.290 ⇒ 00:33:06.710 Uttam Kumaran: A customer, like, a cohesive solution, where, like.
234 00:33:06.980 ⇒ 00:33:20.339 Uttam Kumaran: everyone around them is, like, friends, you know, basically. And so when we come into a client, we’re typically making 5 to 10, like, vendor decisions, right? And so we want to make sure that, like, yes, we’re recommending the best people, but also, it’s, like.
235 00:33:21.070 ⇒ 00:33:34.560 Uttam Kumaran: even… it’s… for us, it’s actually… for me, it’s so much less about, like, I don’t care about the referral fees or anything. For me, it’s actually purely, like, they’re getting a much better experience, because we know that these guys are the best. But I’m also telling them, like, look, they sit in, like, a solution.
236 00:33:34.670 ⇒ 00:33:47.649 Uttam Kumaran: Right? And it’s not as easy, as just, like, plugging and playing. And so, like, that’s sort of how we think about a lot of our vendor partnerships. And again, I’m an engineer, so, like, at minimum, I just want to implement the best tools.
237 00:33:47.650 ⇒ 00:33:59.289 Uttam Kumaran: And, like, because it makes our life, like, way easier. So, even if we don’t have a partnership with folks, we still recommend, like, a set of tools, and we steer them away from bad things, but the more that we can just, like.
238 00:33:59.320 ⇒ 00:34:04.070 Uttam Kumaran: Be friends with those folks, and it helps streamline, like, everybody’s engagement with the customer, so…
239 00:34:04.070 ⇒ 00:34:13.330 Hoshang Mehta: No, 100%. I think service plus solution has always been happening in the data space, and I think even AI doesn’t change that.
240 00:34:13.510 ⇒ 00:34:13.830 Uttam Kumaran: Yeah.
241 00:34:13.830 ⇒ 00:34:29.490 Hoshang Mehta: it kind of needs it more, now. Yeah. So, definitely align with that, and if you have some kind of a form or something that, I can fill up to give more information, I can do that. And,
242 00:34:29.500 ⇒ 00:34:35.009 Hoshang Mehta: Yeah, let’s get together on Slack. We can maybe discuss on how we can structure this partnership as well.
243 00:34:35.010 ⇒ 00:34:35.520 Uttam Kumaran: Sure.
244 00:34:35.520 ⇒ 00:34:46.310 Hoshang Mehta: Because I do have a few clients which I do want to introduce you to in terms of, you know, this is what they do, and if you have any of these issues, we can solve together.
245 00:34:46.500 ⇒ 00:34:47.340 Hoshang Mehta: Fair.
246 00:34:47.610 ⇒ 00:34:50.420 Hoshang Mehta: That’ll make my sales process also faster to grow.
247 00:34:50.429 ⇒ 00:34:51.009 Uttam Kumaran: Totally.
248 00:34:51.520 ⇒ 00:35:00.750 Uttam Kumaran: Yeah, and also anything you need in terms of materials, so we’ll… when we get in Slack, we’ll send you all of our materials. If you want to, like, co-brand stuff, or if you want us to work on, like.
249 00:35:00.750 ⇒ 00:35:11.949 Uttam Kumaran: something that, like, helps you kind of smooth that out, more than happy to. Like, we have design folks on our team that can do that, so we’re happy to take that off your plate. And yeah, even just, like, we can sign a simple…
250 00:35:11.950 ⇒ 00:35:28.280 Uttam Kumaran: like, the easiest thing we can do is just sign a simple agreement that, like, we’re working together, and even… again, I’ll be very upfront, like, I care very less… little about, like, referral fees. It’s, like, not what I’m interested in at all. More, I’m interested in, like, making sure that your clients, like, get up to speed on your tool, and that they…
251 00:35:28.280 ⇒ 00:35:34.979 Uttam Kumaran: get success, right? And so for us, it’s more about, like, can we co-market together? Can we do events together? Like, that’s…
252 00:35:35.040 ⇒ 00:35:50.959 Uttam Kumaran: much more interesting to me than, like, here’s 10%, like, I’m just not interested in that. So that’s what we’re trying to tell people up front, because, like, my currency is like, hey, if you like us, then let’s, like, market together and, like, do more stuff together, you know? Because our clients are facing the same problem, so…
253 00:35:50.960 ⇒ 00:36:05.169 Hoshang Mehta: Yeah, I love doing that kind of a partnership as well. Just want to solve the problem for the client so that, you know, their data is right, my tool gets them to the AI part of it, and the client is happy, so we get more.
254 00:36:05.170 ⇒ 00:36:05.750 Uttam Kumaran: Yeah.
255 00:36:05.750 ⇒ 00:36:07.510 Hoshang Mehta: Anyways, by the way, like, you know, whatever.
256 00:36:07.510 ⇒ 00:36:11.589 Uttam Kumaran: Exactly, it’s just, it could be that easy, you know?
257 00:36:11.590 ⇒ 00:36:19.410 Hoshang Mehta: Yeah, no, that makes sense. Let’s get on Slack, I’ll send you… or if you want to send me a link, I can…
258 00:36:19.410 ⇒ 00:36:20.110 Uttam Kumaran: I will.
259 00:36:20.420 ⇒ 00:36:32.300 Hoshang Mehta: get on that, and yeah, I mean, you can share that agreement, happy to sign that. And we can talk about how we can market together as well, and we can make introductions to the customer side, too.
260 00:36:32.300 ⇒ 00:36:33.700 Uttam Kumaran: Sure, sure, okay, perfect.
261 00:36:33.700 ⇒ 00:36:34.349 Holly Condos: Sounds great.
262 00:36:34.350 ⇒ 00:36:35.730 Hoshang Mehta: Yeah.
263 00:36:36.140 ⇒ 00:36:39.650 Uttam Kumaran: Alright, well, thank you so much for the time, I appreciate it, yeah, and we’ll talk soon.
264 00:36:39.650 ⇒ 00:36:41.400 Hoshang Mehta: Likewise. Thank you so much for your time as well.
265 00:36:41.400 ⇒ 00:36:43.010 Uttam Kumaran: Thanks. Take care. See ya.
266 00:36:43.220 ⇒ 00:36:43.880 Hoshang Mehta: Bye.