Meeting Title: Brainforge AI Tech Lead Interview Date: 2026-03-13 Meeting participants: Vishnu Kakaraparthi, Samuel Roberts


WEBVTT

1 00:02:34.950 00:02:35.620 Samuel Roberts: Hello.

2 00:02:35.620 00:02:37.570 Vishnu Kakaraparthi: Hey, hi. Can you hear me, Sam?

3 00:02:37.570 00:02:45.160 Samuel Roberts: Sorry. Yes, I can… sorry, it’s getting… oh, there we go, hold on. My computer’s a little… here we go. Alright.

4 00:02:45.930 00:02:47.479 Samuel Roberts: How are you today?

5 00:02:47.480 00:02:48.880 Vishnu Kakaraparthi: I’m good, how are you?

6 00:02:50.140 00:03:02.870 Samuel Roberts: Doing alright, doing alright. Thanks for taking the time. So I’ll just jump right in, explain. So, yeah, my name is Sam Roberts. I’m the AI, and Automation Tech Lead here at Brainforge.

7 00:03:03.350 00:03:21.260 Samuel Roberts: I think the way this’ll go is I’ll let you introduce yourself a little bit, and then I have some questions, I’ll make sure we leave time for you to ask any questions, and we’ll just kind of, you know, keep it… keep it like that. So, yeah, if you want to introduce yourself to me, very, you know, briefly, kind of thing.

8 00:03:21.260 00:03:36.369 Vishnu Kakaraparthi: Go ahead. Yeah, I’ll try to keep it concise. Yeah, I’m a PhD student at ASU. I… it’s in computer science, AI, variables, computer vision, a lot of different fields coming through with what we are developing and stuff.

9 00:03:37.760 00:03:52.180 Vishnu Kakaraparthi: since the job is more related to Gen AI and stuff, I do have a lot of experience with that, too, through my past internships and stuff. I worked at Boomi. We developed a documentation engine where

10 00:03:53.000 00:03:59.409 Vishnu Kakaraparthi: they… Boomi actually builds pipelines, and whoever the customer wanted,

11 00:04:00.380 00:04:11.000 Vishnu Kakaraparthi: summaries or detailed reports on the pipeline and things like those. So, the whole idea of this documentation engine was to generate those from the pipelines that they created, or…

12 00:04:11.090 00:04:25.310 Vishnu Kakaraparthi: there is another project that was happening, too, which was automating using Gen AI, these pipelines and stuff, too, so all of that had to come together in one cohesive pattern, where they could upload documents about, like, there are

13 00:04:25.310 00:04:32.470 Vishnu Kakaraparthi: specific company keywords and terms, so we developed the whole documentation engine and stuff like that, for Boomi. Very cool.

14 00:04:32.790 00:04:48.010 Vishnu Kakaraparthi: And yeah, and from my research side, we do a lot of multimodal development. I do a lot of computer vision side of it, but we have a lot of text-based and we have other sensors, too, from…

15 00:04:48.540 00:04:51.279 Vishnu Kakaraparthi: gyroscopes and stuff like that that give us.

16 00:04:51.280 00:04:51.829 Samuel Roberts: Oh, snap.

17 00:04:51.830 00:04:59.650 Vishnu Kakaraparthi: Acceleration data… data and things like… like accelerometer data and all that stuff that we’ve… we’ve… I would…

18 00:04:59.850 00:05:05.490 Vishnu Kakaraparthi: I would be truthful, we still don’t use it completely. We have, like, proof of concepts, it’s all.

19 00:05:05.490 00:05:05.820 Samuel Roberts: Sure.

20 00:05:05.820 00:05:22.379 Vishnu Kakaraparthi: and pieces, it’s not cohesive yet, but the idea is, okay, we can make them cohesive and stuff like that, and things like those, so we identify gestures and things like those from accelerometers. Like, basically, actually, a small, quick explanation of that project is.

21 00:05:22.700 00:05:37.589 Vishnu Kakaraparthi: It’s a camera on a smartwatch. We have it on the dorsal side of… ventral side of the wrist. We have patents with… it could be anywhere, ventral or dorsal and stuff like that, but the idea is we look at what people do with their hands and stuff.

22 00:05:37.690 00:05:43.419 Vishnu Kakaraparthi: I’m actually interested in activities of daily living for older adults living individually and people with dementia.

23 00:05:43.730 00:05:51.440 Vishnu Kakaraparthi: And within that, I’m looking at medication adherence, I’m looking at how people take pills, and is the pill going into their mouth? That’s my, big…

24 00:05:51.910 00:05:59.660 Vishnu Kakaraparthi: research statement that we could track pills going into the mouth, where most of these smart pill boxes and stuff like that.

25 00:05:59.870 00:06:00.970 Vishnu Kakaraparthi: only…

26 00:06:01.170 00:06:19.590 Vishnu Kakaraparthi: go to one extent where they are, like, a reminder, or more or less like, okay, you did… here, here is it, and it’s like a nudge, but this is ensuring that it goes into that. We’re also moving it to smart manufacturing and compliance and things like those, so this technology can be used.

27 00:06:19.650 00:06:26.200 Vishnu Kakaraparthi: different other things, and other projects and internships I worked with, BrainChip.

28 00:06:26.510 00:06:35.049 Vishnu Kakaraparthi: Very much into computer vision, quantized models, and edge… edge.

29 00:06:35.050 00:06:35.560 Samuel Roberts: Cool.

30 00:06:35.560 00:06:40.079 Vishnu Kakaraparthi: Edge deployment and stuff like that, and… yeah,

31 00:06:40.550 00:06:49.349 Vishnu Kakaraparthi: I have an MBA, so I know what… what… like, when there’s a problem, I know how to dissect the problem, what the ROIs are, things like that.

32 00:06:49.350 00:06:50.350 Samuel Roberts: Sure, sure.

33 00:06:50.350 00:06:56.240 Vishnu Kakaraparthi: And when I saw Brain Forge, I saw an interview and stuff like that with Uttam and stuff.

34 00:06:56.360 00:06:59.209 Vishnu Kakaraparthi: And he was literally saying that, like, oh…

35 00:07:00.840 00:07:18.400 Vishnu Kakaraparthi: when you dissect the problem, we know where the problems are, and we can provide solutions. So, that’s very much how I work through any project or anything, too. Like, even with my research, we are on the ground, working with,

36 00:07:18.770 00:07:35.409 Vishnu Kakaraparthi: senior living facilities, we call them, show them the device, ask them what do you like, what do you not like? We had interviews, focus groups, and all that stuff, too. Sometimes they take a lot of time, but then sometimes it’s important to do it, so we need to balance that.

37 00:07:35.410 00:07:40.219 Vishnu Kakaraparthi: So yeah, that’s more or less, basically, my small gist, I guess.

38 00:07:40.220 00:07:50.070 Samuel Roberts: Great, yeah, thank you, that’s really cool. Alright, so we’ll just jump into some of these questions. So, you mentioned that kind of LLM-based feature, but I’m wondering, what…

39 00:07:50.070 00:08:00.420 Samuel Roberts: let’s talk about, like, what part of the AI stack, or Gen AI stack, I guess, have you spent the most time building versus just, like, experimenting and hanging around with? So you mentioned the RAG system a little bit, but I’m wondering.

40 00:08:00.420 00:08:00.910 Vishnu Kakaraparthi: Yeah.

41 00:08:00.910 00:08:01.730 Samuel Roberts: within that.

42 00:08:02.050 00:08:19.680 Vishnu Kakaraparthi: So, yeah, I mean, we started with LangChain, things like those, and then we… we, on our own, developed, like, a router kind of a system. We had two different components. We were doing, like, just semantic routing, where we just have a basic traditional,

43 00:08:20.250 00:08:27.849 Vishnu Kakaraparthi: like, NLP-based router for the question or summary, because the problem was when you give the question

44 00:08:27.850 00:08:41.239 Vishnu Kakaraparthi: the LLM was just answering based on whatever it thinks it should answer. It wasn’t specific to the use case, so we developed different prompt-level stuff for the LLM router, and we also implemented,

45 00:08:41.630 00:08:45.810 Vishnu Kakaraparthi: an LLM-based router system, too, but then…

46 00:08:46.160 00:09:00.439 Vishnu Kakaraparthi: it’s expensive, and it’s a company, it’s a running business, so they don’t want so many API calls. I mean, it’s, again, the decision, is it bringing as much money versus things like yours? So, we have… we developed options, and it’s finally…

47 00:09:00.710 00:09:15.090 Vishnu Kakaraparthi: however they want to deploy it, if the use case becomes much more wide and things like those, because when we were developing, we only had, like, 5, 10 different classes and classifications of the prompt, so…

48 00:09:15.090 00:09:15.630 Samuel Roberts: Okay.

49 00:09:15.630 00:09:19.439 Vishnu Kakaraparthi: It wasn’t that bad, to develop these

50 00:09:20.170 00:09:29.939 Vishnu Kakaraparthi: the basic NLP traditional stuff and things like those. Other than that, I have done some, like, not with

51 00:09:30.000 00:09:41.049 Vishnu Kakaraparthi: any company or something like that, but I’ve done some pre-training of models and stuff like that, and with my research, too, I have, pre- like, trained

52 00:09:41.170 00:09:47.589 Vishnu Kakaraparthi: large, large vision models and stuff like that. I use VIVIT, so that’s, like, a huge, big…

53 00:09:47.710 00:09:54.480 Vishnu Kakaraparthi: A transformer-based, video-based transformer, model, so, yeah.

54 00:09:54.580 00:10:05.860 Vishnu Kakaraparthi: I do have experience with that. Yeah, I keep diving into everything. Today, I was just reading about the Google’s embedding, too, which is their multimodal embedder and stuff like that. So, yeah, I mean…

55 00:10:06.320 00:10:08.770 Samuel Roberts: Yeah. Okay, let’s, let’s,

56 00:10:09.520 00:10:22.559 Samuel Roberts: That’s actually kind of a good segue, because there’s so much stuff coming out, and there’s lots of, changes in the industry, everything, there’s a new model, a new framework, a new this, a new that. When has there been something that…

57 00:10:22.600 00:10:41.699 Samuel Roberts: and this might be a little different, I suppose, depending on, like, your work for your PhD versus, like, for a business, but some kind of trend in Gen AI that you were excited about, and maybe tried out, but decided not to adopt for whatever project because it wasn’t ready, or because it wasn’t right, or… can you talk to me a little bit about that thought process?

58 00:10:44.570 00:10:57.510 Vishnu Kakaraparthi: It would be probably the new models that keep coming up every day. I would probably say I generally stick with one model, because in my head, I can see the improvements of the other things that I’m trying to do.

59 00:10:57.850 00:11:07.189 Vishnu Kakaraparthi: and keep track that, okay, the other modules that I’m working on are improving the scores better than just the model itself. I don’t really trust

60 00:11:07.520 00:11:16.449 Vishnu Kakaraparthi: I wouldn’t say trust is the correct word, but I don’t really think the emphasis is the actual model. It’s about how you integrate the model with other things.

61 00:11:16.450 00:11:17.080 Samuel Roberts: Sure.

62 00:11:17.080 00:11:34.139 Vishnu Kakaraparthi: just basic, like, simple tasks and things like those, which generally are, in everyday life, we are using LLMs for. We don’t need these huge, big models, so my thing… my whole understanding is quantize, small, can we do it with

63 00:11:34.530 00:11:38.910 Vishnu Kakaraparthi: simple model, that’s the thing that I do, so I would say…

64 00:11:39.070 00:11:41.370 Vishnu Kakaraparthi: Not keeping up with the trend, or, like, trying…

65 00:11:41.370 00:11:41.750 Samuel Roberts: Yeah.

66 00:11:41.750 00:12:01.149 Vishnu Kakaraparthi: get into that wormhole of, oh, we have the new biggest model, the new multi-billion dollar spent model, or whatever is… I like playing with it, I tinker around, I see the scores, I see what it’s doing, but I just don’t download the latest model every day and run the…

67 00:12:01.220 00:12:04.739 Vishnu Kakaraparthi: project, I guess, on my… on the latest model.

68 00:12:05.050 00:12:08.760 Samuel Roberts: Makes sense, makes sense. Alright, let’s talk a little bit about,

69 00:12:09.010 00:12:32.569 Samuel Roberts: so all this news out there about all this Jet AI stuff, we deal with a lot of clients, so the way we kind of work, and I’ll explain this more later, that’s probably, but, you know, we have client work, we also have, like, some internal work, and so, that is different because we’re building stuff for ourselves, but when we’re using… when we’re doing stuff for clients, not only do we have to figure out what is production-ready, but also communication with clients and non-technical stakeholders.

70 00:12:32.570 00:12:33.510 Samuel Roberts: who…

71 00:12:33.530 00:12:47.560 Samuel Roberts: sometimes might think they know what’s possible because they hear all this stuff, and they’re following the news. So, how do you go about explaining the limitations of some of this technology to a non-technical stakeholder? Like, what are the things you think about there?

72 00:12:48.100 00:12:49.799 Vishnu Kakaraparthi: I would generally first…

73 00:12:50.310 00:13:03.369 Vishnu Kakaraparthi: probably the starting point with any project is probably discuss what they really want, right? When you put that in words, when you put… see things clearly that, okay, this is what happens. Then I can also tell them about limitations, like.

74 00:13:03.370 00:13:14.610 Vishnu Kakaraparthi: It might not work 100% every time. There probably will be issues. There’ll be, sometimes it’ll answer wrong. I’ll explain them what hallucinations are. It could hallucinate, or someone could even…

75 00:13:14.610 00:13:32.670 Vishnu Kakaraparthi: hack it into going into a loop and work around things and stuff like that. There will be people who’ll try to exploit, and I’ll tell them what resources they have to address those, or what resources it might cost us as a company or as a client.

76 00:13:32.670 00:13:36.260 Vishnu Kakaraparthi: As we are onboarding them, what could…

77 00:13:36.410 00:13:42.970 Vishnu Kakaraparthi: be the problems and issues, and also possible solutions for those.

78 00:13:43.100 00:13:48.920 Vishnu Kakaraparthi: I don’t know if I answer… answered your question with that, but yeah, that’s more or less, like.

79 00:13:49.030 00:13:53.509 Vishnu Kakaraparthi: it’s more communicative, and it’s iterative, and I like…

80 00:13:53.750 00:14:09.899 Vishnu Kakaraparthi: like I always… I was saying before, too, I like the client, or whoever we are developing this technology for in the chair, so we send them updates, or give them, what is happening right now, or what is the current,

81 00:14:10.030 00:14:15.260 Vishnu Kakaraparthi: worldscape, I guess, about… The…

82 00:14:15.640 00:14:21.820 Vishnu Kakaraparthi: about their use case, and what is… what are other people doing, and what are we doing, and how, like, it’s more about…

83 00:14:22.150 00:14:23.190 Vishnu Kakaraparthi: keeping it…

84 00:14:24.800 00:14:33.509 Vishnu Kakaraparthi: organized and clear to them that what goals are and what you should… what they should expect out of it. Like, that’s probably it.

85 00:14:33.740 00:14:37.549 Vishnu Kakaraparthi: Cool. Thank you. Yeah.

86 00:14:37.550 00:14:39.989 Samuel Roberts: Trying to think where else we would dig in a little bit here.

87 00:14:41.690 00:14:47.650 Samuel Roberts: So, maybe digging on that a little bit, has there been a time when… A user misunderstood what

88 00:14:47.830 00:14:52.920 Samuel Roberts: a feature was doing, or possibly could do. Is there a time when, you know.

89 00:14:53.020 00:15:02.289 Samuel Roberts: Something underperformed, and you had to figure out why, and communicate that, and, you know, talk to me a little bit about, kind of, when problems arise that way.

90 00:15:02.440 00:15:21.080 Vishnu Kakaraparthi: Yeah, for sure. Like, one of the projects I worked with is a company called Movement Interactive. It’s also related to senior living facilities and stuff like that. They were building a device with accelerometers for fall detection, because fall… fall is the biggest issue for them, so that they can track someone’s

91 00:15:21.400 00:15:27.249 Vishnu Kakaraparthi: has fallen, or something like that. The problem is the false positives are very, like.

92 00:15:27.470 00:15:31.120 Vishnu Kakaraparthi: Very few people fall, and we don’t have real data, and there’s…

93 00:15:31.120 00:15:31.550 Samuel Roberts: You get.

94 00:15:31.550 00:15:44.029 Vishnu Kakaraparthi: a lot of false positives, because the AI wants to just predict the… even at, like, lowercase, it wants to just predict, oh, someone fell. So that scenarios are biggest issues we face.

95 00:15:44.320 00:16:03.029 Vishnu Kakaraparthi: And my solution for that was I wanted to do… I mean, we tried doing it, we didn’t complete the whole project, I just gave them, like, a simple proof of concept of how that works, but then the idea was to do something called post-fall analysis, see and understand stuff, what

96 00:16:03.030 00:16:05.960 Vishnu Kakaraparthi: generally people do. Oh, after fall, like.

97 00:16:05.960 00:16:19.329 Vishnu Kakaraparthi: are they stuck? They’re not moving anywhere. Or did they get up and walk? Or did they… are they rolling around? Things like those are different scenarios that can happen after fall, so…

98 00:16:19.330 00:16:27.569 Vishnu Kakaraparthi: we calculated a little data out of that, and I trained another model with the accelerometer data and saw that. So…

99 00:16:28.200 00:16:40.510 Vishnu Kakaraparthi: Like, again, I was answering the question, yes, we do have… we do identify some problems where the client will not probably… I mean, the client was the company, but there it is, that…

100 00:16:41.190 00:16:46.540 Vishnu Kakaraparthi: We had to solve that problem there by introducing something else, and…

101 00:16:46.820 00:16:57.060 Vishnu Kakaraparthi: follow through with that? Do I get a better result out of it, just not throwing in one model that would just be, like, just fault detection and stuff like that?

102 00:16:57.060 00:16:58.819 Samuel Roberts: Cool, that’s good, thank you.

103 00:16:59.030 00:17:12.019 Samuel Roberts: Alright, so we’re almost halfway. I kind of wanted to make sure we have time for you to ask any questions. I don’t want to just get to the end and not have enough time. So, if there’s any questions you have for me about the role, Brainforge, whatever I can answer, I’m happy to…

104 00:17:12.250 00:17:28.679 Vishnu Kakaraparthi: Yeah, surely, can you just, like, there wasn’t that much information about the role, like, at what, yeah, what stage is the company at? I… I heard, from Utam’s interviews, you said there, he said he… there were 15 people, is that the current size?

105 00:17:28.680 00:17:33.030 Vishnu Kakaraparthi: Of the company, and how many, like, clients do you onboard, and do you have in general?

106 00:17:33.030 00:17:34.030 Samuel Roberts: Yeah…

107 00:17:34.030 00:17:35.150 Vishnu Kakaraparthi: Things like those.

108 00:17:35.410 00:17:45.549 Samuel Roberts: Yeah, I mean, 15 is maybe a little low now. We’ve been growing a bit. Obviously, like, we’re hiring, so the more people have been coming on, we’ve been doing, different

109 00:17:45.710 00:17:48.810 Samuel Roberts: parts of the company. So, just a little…

110 00:17:49.130 00:17:55.039 Samuel Roberts: context, I guess. So the engineering side of the company is kind of split into two teams. There’s the data side.

111 00:17:55.040 00:18:10.130 Samuel Roberts: And so, they handle ETLs and pipelines and modeling and all this other stuff for certain clients that have that kind of work. And then the AI side kind of spun out of that and the internal work that we were doing, because we started as a data consultancy.

112 00:18:10.130 00:18:15.460 Samuel Roberts: And then started adding other services. And so we have about,

113 00:18:15.460 00:18:33.749 Samuel Roberts: I don’t know the exact number of people, but on the AI team, there’s me, and there’s three other engineers, and we’re working on… so we have kind of fewer AI clients than data clients at this point, just because of the nature of how things started. But some of those data clients become AI clients, and other AI clients come in. So, at any given time, we have

114 00:18:33.750 00:18:37.319 Samuel Roberts: You know, 4 of us were looking to add a fifth engineer.

115 00:18:37.400 00:18:51.290 Samuel Roberts: And so, you know, we kind of… clients kind of come and go, depending on how big the projects are. We have one right now that we’re working on that is a similar kind of RAG system. It’s a customer service

116 00:18:51.800 00:19:04.540 Samuel Roberts: chatbot for the customer service representatives, not the actual end users. So, this company had a lot of data everywhere, and we’ve been working on that for a while in different forms and refining it.

117 00:19:05.370 00:19:12.159 Samuel Roberts: Yeah, so I mean, we’re… we have a very, kind of, startup-y culture, you know, we’re a consultancy, but it’s not a,

118 00:19:12.360 00:19:16.300 Samuel Roberts: You know, we’re not a product company, so it’s, a little different.

119 00:19:18.540 00:19:25.620 Samuel Roberts: But… yeah, so I actually… yeah, I want to let you keep asking questions, but I have some other questions after that, but go ahead, yeah.

120 00:19:25.730 00:19:26.400 Samuel Roberts: Yeah, I’m sorry.

121 00:19:26.400 00:19:32.589 Vishnu Kakaraparthi: So, technically, you guys work with one or two clients, is generally the norm, is what you’re saying, or…

122 00:19:32.590 00:19:37.799 Samuel Roberts: I think at this size, we’ve done… we’ve done… we’ve had up to…

123 00:19:37.970 00:19:40.470 Samuel Roberts: like, 4 AI projects, I think, at once.

124 00:19:40.470 00:19:41.200 Vishnu Kakaraparthi: Okay.

125 00:19:41.200 00:19:57.089 Samuel Roberts: We also do tech… internal work, like I mentioned, so we’re doing internal tooling and platforms and stuff, so that’s kind of… we think of that as, like, a separate client, because it’s allocating. But yeah, and then I would say from there, you know, we’re looking to sell more right now, that’s why we’re looking to bring on more people.

126 00:19:57.120 00:20:02.560 Samuel Roberts: We’re kind of hitting our capacity with some of this stuff, as we have, but yeah.

127 00:20:03.120 00:20:04.370 Vishnu Kakaraparthi: Right.

128 00:20:04.490 00:20:08.099 Vishnu Kakaraparthi: That generally answers, like, I just wanted to know the scope of.

129 00:20:08.200 00:20:08.790 Samuel Roberts: Yeah.

130 00:20:08.790 00:20:11.309 Vishnu Kakaraparthi: any projects I work on and stuff. I do like…

131 00:20:11.990 00:20:18.079 Vishnu Kakaraparthi: the whole idea of selecting a company like you is, like, I do like to work on multiple projects, that’s my.

132 00:20:18.080 00:20:18.490 Samuel Roberts: Yeah.

133 00:20:18.490 00:20:20.300 Vishnu Kakaraparthi: jam,

134 00:20:20.480 00:20:27.439 Vishnu Kakaraparthi: focusing on just one project sometimes gets tedious. It’s so much easier to work on two projects or three projects.

135 00:20:27.440 00:20:28.070 Samuel Roberts: I understand.

136 00:20:28.070 00:20:32.619 Vishnu Kakaraparthi: working on one, you’re like… you feel like you’re stuck on one, you’re like, okay, let’s do the other one.

137 00:20:32.620 00:20:33.060 Samuel Roberts: Yes.

138 00:20:33.060 00:20:33.660 Vishnu Kakaraparthi: I’m tired.

139 00:20:33.660 00:20:37.590 Samuel Roberts: It’s interesting, because sometimes contact switching is bad, and sometimes contact switching is good.

140 00:20:37.590 00:20:50.639 Vishnu Kakaraparthi: Yeah, yeah, so that’s… I… I did interview with another company that… they’re very similar to you, but they are, like, more venture capitalistic. They… they acquire companies that.

141 00:20:51.060 00:21:04.489 Vishnu Kakaraparthi: they think have good positive revenue and stuff, and they add AI and all these features, and then they, sell it off, technically. That’s their more… more or less their business plan. They do have a consultancy side, but I think that, according to

142 00:21:04.610 00:21:09.020 Vishnu Kakaraparthi: talking to them, I think they make more money this way than consulting and stuff.

143 00:21:09.020 00:21:09.469 Samuel Roberts: Okay, yeah.

144 00:21:09.470 00:21:17.089 Vishnu Kakaraparthi: Especially since… I mean, the venture capitalist, pie is very big right now, so…

145 00:21:17.090 00:21:19.549 Samuel Roberts: Yes. Yeah, so… It is, it is, yeah.

146 00:21:19.550 00:21:20.130 Vishnu Kakaraparthi: So.

147 00:21:20.130 00:21:20.570 Samuel Roberts: Yeah.

148 00:21:20.760 00:21:30.799 Vishnu Kakaraparthi: That’s what they do, so I’m, like… that’s why I’m, like, looking at companies which do these, like, multiple small projects and have, like, multiple clients and things like that. That’s what interests me.

149 00:21:30.800 00:21:46.410 Samuel Roberts: Yeah. No, it’s interesting, because I come from more of a startup product background, and so this has been a little bit of a switch for me, where I’m used to working on, you know, one main product, obviously lots of features, but the different problems we’re solving in different ways, and, you know, seeing where the patterns are, and what we can…

150 00:21:46.550 00:21:53.239 Samuel Roberts: pull out of those projects and maybe productize into something that we can more easily sell as a service line. It’s actually really interesting stuff.

151 00:21:53.980 00:21:54.530 Vishnu Kakaraparthi: phone.

152 00:21:55.190 00:21:58.190 Samuel Roberts: That triggered a thought, and I don’t remember what it was now.

153 00:22:00.020 00:22:07.510 Samuel Roberts: Oh, yeah, okay, so I was just gonna say, yeah, so we have a few different clients at any given time, you’ll be bouncing around, probably between one or two of those, and some internal stuff.

154 00:22:07.660 00:22:12.840 Samuel Roberts: Obviously, things change, and allocations change over time, so you never know, but, yeah.

155 00:22:12.840 00:22:13.700 Vishnu Kakaraparthi: No worries.

156 00:22:13.700 00:22:15.099 Samuel Roberts: Other… other questions, or…

157 00:22:15.100 00:22:19.399 Vishnu Kakaraparthi: I have questions about stage 2 of the interview, but we can do that at the end.

158 00:22:19.400 00:22:34.109 Samuel Roberts: Yeah, I’ll explain all that, but… Yeah, I guess, the last thing I want to ask a little bit about, so I mentioned there’s, sort of four of us right now on the AI team, kind of specifically. I’m curious about your experience working on a team, I don’t know.

159 00:22:34.130 00:22:42.249 Samuel Roberts: how things are different in academia versus in business. Like, I’m just curious, like, where… how that fits in there, you know. I came from a very, like,

160 00:22:42.350 00:22:48.370 Samuel Roberts: small startup environment where it was, like, me and a few other people, so small teams were really how I kind of grew. I’m curious.

161 00:22:48.370 00:22:54.380 Vishnu Kakaraparthi: more or less, it was always small teams that I worked with. Ericsson.

162 00:22:54.380 00:23:09.970 Vishnu Kakaraparthi: me and two other people, so one at my same level and technically our boss, I would say. So three, they don’t… they don’t like calling people in, like, a boss. They’re, like, mentors or whatever, like, whatever, but then, yeah.

163 00:23:09.970 00:23:20.510 Vishnu Kakaraparthi: So, just 3 people when I worked there. BrainChip was… they had a bigger team, but I only interacted with, like, 4 to 5 people on every given week.

164 00:23:20.670 00:23:34.709 Vishnu Kakaraparthi: in academia, it’s more… technically, fair to say, it’s more lonelier. I have a hardware guy who I… I’m not a big hardware person, like, I can solder stuff, I know, but I cannot develop anything new.

165 00:23:34.710 00:23:35.359 Samuel Roberts: It’s like…

166 00:23:35.360 00:23:36.400 Vishnu Kakaraparthi: else,

167 00:23:36.510 00:23:46.650 Vishnu Kakaraparthi: I have a hardware guy, we only interact with him, and all the others are professors, so they are either AI professors or human-computer interaction professors, and…

168 00:23:46.850 00:23:53.529 Vishnu Kakaraparthi: Things like those. They’re bringing different views and viewpoints and things like those, that’s more or less, yeah.

169 00:23:54.100 00:23:55.300 Samuel Roberts: Cool, alright, good, thank you.

170 00:23:55.300 00:23:58.539 Vishnu Kakaraparthi: I don’t think I’ve worked with more than 5 or 10 people.

171 00:23:58.540 00:23:58.870 Samuel Roberts: Okay.

172 00:23:58.870 00:24:00.590 Vishnu Kakaraparthi: I’m in a project.

173 00:24:00.590 00:24:08.410 Samuel Roberts: Yeah, and that matches pretty well, then. Alright, I think that’s pretty much all I have. I’ll,

174 00:24:08.570 00:24:22.609 Samuel Roberts: I guess I’ll explain the rest of the process. So, I’ll bring this back to the team, and then the next step, if you pass kind of this check, is a second interview that’s a little more role-focused, a little more technical.

175 00:24:22.800 00:24:27.529 Samuel Roberts: And then, if you pass that one, there’s a little bit of a tech assessment.

176 00:24:27.890 00:24:47.000 Samuel Roberts: That we would give you, and then we would do a, a panel interview after that, where we’d talk about that, ask you questions, and, about the implementation and stuff like that. And then after that would be an offer, and then, starting. So, kind of a three-phase, process. We like to move relatively quickly.

177 00:24:47.000 00:24:58.260 Samuel Roberts: I think the biggest thing is just scheduling the time with people, you know, so, you know, you should hear back pretty quickly, after this, I imagine, one way or another. And then,

178 00:24:58.810 00:25:01.259 Samuel Roberts: Yeah, that’s basically the whole process.

179 00:25:01.900 00:25:14.379 Vishnu Kakaraparthi: So, is the technical assessment part of Stage 2? Because that’s what I read. It says deeper discussion and practical evaluation. Is that ad hoc, or what do you expect from practical evaluation standards?

180 00:25:14.380 00:25:25.430 Samuel Roberts: The practical evaluation… so the next interview, like I said, is kind of focused on the role, a little more technical. You’ll meet with one of the other engineers, and he’ll ask some more questions that are less…

181 00:25:26.060 00:25:26.780 Samuel Roberts: I don’t know.

182 00:25:27.160 00:25:39.860 Samuel Roberts: abstract, maybe then, yeah, exactly. And then I believe after that is when you would have the actual assessment. We’d send you, I think it’s just a GitHub repo and some instructions there for

183 00:25:40.180 00:25:45.439 Samuel Roberts: how to submit a project, and then that would be the panel discussion, I think, after that.

184 00:25:46.240 00:25:59.739 Vishnu Kakaraparthi: Sounds great, then. Now, yeah, I just wanted to make sure I’m in the right space for the next stage, when… is… is it happening on the interview date, or is it happening as text assessment happens after the stuff was…

185 00:26:00.040 00:26:01.549 Vishnu Kakaraparthi: Well, that sounds great, Ben.

186 00:26:01.680 00:26:02.780 Vishnu Kakaraparthi: Alright, cool.

187 00:26:02.780 00:26:08.220 Samuel Roberts: Then I think if that’s… if you don’t have any other questions, I think I’m… I don’t have any other questions either, so…

188 00:26:08.230 00:26:11.020 Vishnu Kakaraparthi: Awesome! Thank you so much for meeting. It was great to be.

189 00:26:11.020 00:26:13.100 Samuel Roberts: Yeah, appreciate the time. Great meeting you.

190 00:26:13.100 00:26:16.000 Vishnu Kakaraparthi: Always. Hope to see you again. Bye, see you.