Meeting Title: Brainforge AI Engineer Interview Date: 2026-03-12 Meeting participants: Kaela Gallagher, Nevan Zade
WEBVTT
1 00:01:45.280 ⇒ 00:01:46.330 Kaela Gallagher: Hi.
2 00:01:46.330 ⇒ 00:01:47.170 Nevan Zade: an oven?
3 00:01:48.610 ⇒ 00:01:50.010 Nevan Zade: Hey, Kata, how are you?
4 00:01:50.250 ⇒ 00:01:51.270 Kaela Gallagher: Hi!
5 00:01:52.540 ⇒ 00:01:55.210 Nevan Zade: I’m good, I’m good. Sorry, my Zoom wasn’t working.
6 00:01:55.670 ⇒ 00:01:56.869 Nevan Zade: I have to update that.
7 00:01:57.820 ⇒ 00:01:58.410 Nevan Zade: Nope.
8 00:01:59.180 ⇒ 00:02:00.250 Nevan Zade: Can you see me?
9 00:02:00.250 ⇒ 00:02:03.219 Kaela Gallagher: Yeah, yeah, I can see you just fine. Is it Nevin?
10 00:02:04.040 ⇒ 00:02:05.060 Nevan Zade: Yep, it’s Nemon.
11 00:02:05.060 ⇒ 00:02:20.190 Kaela Gallagher: Nevin, nice to meet you. Thanks for taking some time for me. Yeah, excited to get to know, like, a little bit more about you, and yeah, I guess I would love to just start off by asking, like, what is putting you on the market for a new position?
12 00:02:21.770 ⇒ 00:02:35.170 Nevan Zade: I think the first and the foremost, like, I have good experience in the full-stack development side, but I’m looking more into the AI side of things, because the recent projects I’ve been working on are more inclined towards the AI side, but because of my current role.
13 00:02:35.230 ⇒ 00:02:43.569 Nevan Zade: It’s not easy to switch totally towards AI sites, so I think that’s… that’s something that’s putting me to… to get to look new opportunities.
14 00:02:44.290 ⇒ 00:02:47.139 Kaela Gallagher: Okay, okay. And what are you doing?
15 00:02:47.420 ⇒ 00:02:49.539 Kaela Gallagher: Right now, in your current role?
16 00:02:50.610 ⇒ 00:03:04.460 Nevan Zade: So, right now, I am working as a senior software developer for Valeris Group, and in terms of my job responsibilities, I’m managing a team of 12 individuals. They come on and off based on the project’s requirements.
17 00:03:04.570 ⇒ 00:03:24.059 Nevan Zade: And, currently, I’m dealing with projects related to AI and ML, but I am more into the capacity of integration sides of things, rather than directly working on AI or the machine learning side. So, for example, one other client was in the C-sharp environment, so they had their own prem setup, so I helped them integrate AI with their systems, and
18 00:03:24.060 ⇒ 00:03:31.219 Nevan Zade: In the end, we were doing deployment for them in the Windows environment, so that we have to develop services for them that they can run on Windows.
19 00:03:31.220 ⇒ 00:03:44.660 Nevan Zade: So these sort of work currently going on. So, mainly assisting the AI and ML engineers to deploy their solutions, to actually ship their solutions quickly to the client, and anything related to infrastructure, anything related to,
20 00:03:44.660 ⇒ 00:03:52.509 Nevan Zade: latencies and their, their whole architecture, contracts that they have to define to… between different services, so that’s how to work.
21 00:03:52.950 ⇒ 00:04:01.459 Kaela Gallagher: Okay, so in terms of the way that you were touching AI and ML, it was mostly just, like, partnering with those teams to help them deliver the solutions?
22 00:04:02.410 ⇒ 00:04:21.650 Nevan Zade: is… it is just… this is not exactly around partnering. I have myself developed AI solutions for a client as well. So, mostly the job responsibilities, because I’m a senior software engineer, so I’m coming in as a capacity of a developer rather than an AI engineer myself. But I do have good expertise on the AI side as well.
23 00:04:21.649 ⇒ 00:04:24.759 Nevan Zade: For example, if you’re trying to deploy an LLM,
24 00:04:24.760 ⇒ 00:04:47.809 Nevan Zade: onto, onto any microservice or any other cloud-based service. So you need to have a very, very solid, inference engine, built on top of it. So I have done that. I have deployed, models myself using VLLM, and there needs to be a tokenization flow on top of it. So there was a lot of, work to be done, such as caching and, KV caching.
25 00:04:47.810 ⇒ 00:05:00.280 Nevan Zade: batching and pooling on top of it, so that it can give inference at runtime, so you can get the streaming experience from it. So yeah, I have good experience with that as well, and also I have developed machine learning models in my previous experience as well.
26 00:05:00.280 ⇒ 00:05:09.330 Nevan Zade: But that was very early on with Volaris Group, but later on, mostly it has been towards integrating AI with other systems. And, given that, I guess.
27 00:05:09.370 ⇒ 00:05:16.990 Nevan Zade: from the time onward, like, when the GPT came in, so everything has switched more towards AI rather than machine learning side of things, so it’s more…
28 00:05:17.020 ⇒ 00:05:32.220 Nevan Zade: orchestration side of things, than to… than to work on the actual model development itself. Like, there are very few… few companies in the market that are actually developing models that are… are dedicated to a specific niche. So, yeah, I think,
29 00:05:32.220 ⇒ 00:05:42.909 Nevan Zade: the current need in the market is to enable organizations to actually adopt AI into their system. So I think that’s… that’s where I want to be in the next one or two years.
30 00:05:43.290 ⇒ 00:06:02.700 Kaela Gallagher: Okay, okay, got it. And I’m not super, super technical myself, so, apologies if I’m, like, repeating questions here, but the AI solution, like, you mentioned you built an AI solution for one of your clients. Just, like, very high level, like, what was the purpose of the solution? Like, what did it provide to the client?
31 00:06:03.760 ⇒ 00:06:21.349 Nevan Zade: So, the client itself was in the legal space, so they had a lot of data on legal cases, and it was going towards medical side of things as well. So, for example, if there is some litigation-based case, or some car accident, you need to have some demands put in towards the insurance companies.
32 00:06:21.350 ⇒ 00:06:29.469 Nevan Zade: So they had a lot of information on that, so I have to jump in to develop a solution so that they can automate different processes within their system.
33 00:06:29.640 ⇒ 00:06:42.959 Nevan Zade: So for that, the AI engineers would be jumping in later, but for that specific use case, I have to enable the organization to let it host different AI-based applications into their ecosystem.
34 00:06:42.960 ⇒ 00:06:56.180 Nevan Zade: So I have to develop this microservice-based architecture so that they can develop services that can be hosted within Windows. And the first product that I did with them was to automate the demand letter generation of it.
35 00:06:56.300 ⇒ 00:06:57.370 Nevan Zade: Because…
36 00:06:57.460 ⇒ 00:07:12.180 Nevan Zade: the system itself was pretty isolated. There was… there were a lot of silos in them, so it was, like, 18 to 19 years old system. So, there were a lot of silos. So first, AI needs to be able to access all the right information. For that, I have to develop a rag pipeline.
37 00:07:12.180 ⇒ 00:07:22.600 Nevan Zade: on top of their, their DBs, so that it can pick up information from their DBs, by… on the basis of queries that we can dump into the, into their databases.
38 00:07:22.600 ⇒ 00:07:46.130 Nevan Zade: And the RAG output would then be going towards the LLM, and the LLM would then be outputting the demand letter for the client. And because the demand letters are actually something that cannot be… that cannot be inaccurate, because otherwise the whole claim itself would go down, so you won’t be getting anything out of the insurance companies otherwise. So there was a lot of front engineering into it, and some Jinja template-based
39 00:07:46.130 ⇒ 00:08:05.320 Nevan Zade: rules so that it cannot go wrong beyond the facts that we have. So an evaluation pipeline on top of it, so that it doesn’t get away from the actual purpose of it. So there was this verifier agent on top of it that would verify if the information coming in from the retrieval service, and what the AI is putting forward are accurate and up to the mark.
40 00:08:05.320 ⇒ 00:08:18.379 Nevan Zade: So that was the whole orchestration. So I use Langchain and Langgraph for that. So Langgraph specifically for the agentic orchestration side, and Langchain for doing the RAG-based, retrieval part of it.
41 00:08:18.380 ⇒ 00:08:29.800 Nevan Zade: So that was mostly it, and everything that we built was in Python, so it was Django at the backend, and we have to develop front-end for them as well, and later on, we integrated that with the iframe part.
42 00:08:29.800 ⇒ 00:08:34.830 Nevan Zade: Because the system itself was C-sharp-based, so we have to come up with a way to
43 00:08:34.830 ⇒ 00:08:49.740 Nevan Zade: to actually streamline or lean their development lifecycle, so that they can have this application running on them. So the iframe was the obvious solution for them. So we integrated the iframe. That can be made part of any of their software pages.
44 00:08:49.900 ⇒ 00:09:09.039 Nevan Zade: Yeah, that was mostly it, and in terms of AI, I was building their rack solution, and later on, the demand letter generation. Then later on, the AI developer went in, and there were other solutions, for example, medical technology generation was one of it, that they were working on, so I went in to deploy that as well. So yeah, that sort of…
45 00:09:09.340 ⇒ 00:09:10.889 Nevan Zade: Work is going on on and off.
46 00:09:11.520 ⇒ 00:09:28.490 Kaela Gallagher: Okay, okay, got it. I know you, like, submitted an application for our AI engineer position. Just to give an overview of kind of what we’re looking for right now, we’re bringing on both, like, an AI engineer and a data engineer,
47 00:09:28.570 ⇒ 00:09:48.070 Kaela Gallagher: We, given that we’re, you know, kind of a smaller organization, and we’re very client-facing, we, you know, ask our engineers to be working on multiple projects at one time, we need our engineers to be really good at context switching, but then we also need our engineers to be really active in, like, partnering with our clients.
48 00:09:48.070 ⇒ 00:09:53.340 Kaela Gallagher: And being client-facing, joining meetings, being able to explain really technical things.
49 00:09:53.340 ⇒ 00:09:57.040 Kaela Gallagher: In a way that maybe makes sense to non-technical stakeholders.
50 00:09:57.150 ⇒ 00:10:06.509 Kaela Gallagher: I guess, like, given that, I’m curious if you, like, interact directly with clients in your current position.
51 00:10:07.690 ⇒ 00:10:18.139 Nevan Zade: Yep, most of the time, I’m directly client-facing, so because the devs underneath me, they would jump in when the initial POC is over, and we get the client to see the actual value they can get.
52 00:10:18.140 ⇒ 00:10:34.840 Nevan Zade: From the POC. So then the devs jump in. So, from then point, from then point beyond, beyond this point, it is the dev that is doing the most of the work, and I’m only engaged towards the client side when we have to explain anything to the client, and maybe, there’s a change in requirements and everything.
53 00:10:34.840 ⇒ 00:10:35.880 Nevan Zade: So…
54 00:10:35.880 ⇒ 00:10:59.760 Nevan Zade: So from the point when the problem statement is given to us by the client, I am engaged with the client on the technical front, and there is this product team with us that will be the non-technical people that bridge the gap between us and the developers and the client. So they are the ones who get to the problem themselves, the product guys, and I translate that specific product into a technical
55 00:10:59.760 ⇒ 00:11:12.200 Nevan Zade: roadmap that the dev team can work on. I separate the task for them, and if anything non-technical… anything technical comes in that I need the client to know, or maybe the client wants some explanation on top of it, so I’m mostly involved in that as well.
56 00:11:12.410 ⇒ 00:11:16.959 Nevan Zade: So, yeah, I think… I think Volaris is a consultancy group, so… so they are.
57 00:11:16.960 ⇒ 00:11:17.380 Kaela Gallagher: beautiful.
58 00:11:17.740 ⇒ 00:11:30.510 Nevan Zade: So mostly I’ve been working on contracts related to jobs, like, ranging from 3 months to 12 months. So, so there are on and off assignments coming in and coming out, so yeah, that totally resonates.
59 00:11:31.130 ⇒ 00:11:39.440 Kaela Gallagher: Okay, okay, cool. I know we just have a couple minutes left, so just, curious if you have any questions for me that I can help answer.
60 00:11:40.390 ⇒ 00:11:50.130 Nevan Zade: Sounds good, sounds good. I think, yeah, given that the job description itself is around AI and ML, but there is a mention of Node.js and React.js, so…
61 00:11:50.220 ⇒ 00:12:03.379 Nevan Zade: front-end and back-end-based texts as well. So I wanted to know how much, like, what’s the actual mix there? Is it, like, 80%, 90% AI, 10% of that? So… so how… how does the role entail, like…
62 00:12:03.820 ⇒ 00:12:06.380 Nevan Zade: What is the actual miss in terms of technologies?
63 00:12:06.380 ⇒ 00:12:16.680 Kaela Gallagher: Yeah, yeah, so a lot of our engineers, like you mentioned, come with either, like, a TypeScript or a Python background,
64 00:12:17.160 ⇒ 00:12:28.720 Kaela Gallagher: And then, in terms of, like, the position, so we have… we have a few different work streams. Most people in our company are either on, like, our strategy team, our data team, or our AI team.
65 00:12:28.720 ⇒ 00:12:37.759 Kaela Gallagher: A lot of times, like, our data team would go in with a client first, help sort out the data, and then the AI team can come in and kind of build on that.
66 00:12:37.920 ⇒ 00:12:41.210 Kaela Gallagher: And…
67 00:12:41.740 ⇒ 00:12:48.650 Kaela Gallagher: Yeah, I… in terms of, like, what our AI team is doing, we’re not always necessarily, like, building
68 00:12:48.850 ⇒ 00:12:54.949 Kaela Gallagher: new LLMs, sometimes we’re, like, implementing models that already exist.
69 00:12:55.220 ⇒ 00:13:06.089 Kaela Gallagher: So, yeah, it’s kind of a mix of things. Like, you’re working on multiple clients at a time, and each client, you know, needs something different. So, just kind of depends what client you’re supporting.
70 00:13:06.810 ⇒ 00:13:24.119 Nevan Zade: Sounds good. It feels like we hop onto the client for the end-to-end cycle for them. So I think, yeah, it pretty much resonates with the Microsoft job description as well. And, in terms of the AI and ML solutions, like, what’s the… what’s, like, I don’t know if that’s the right question for this round.
71 00:13:24.130 ⇒ 00:13:36.429 Nevan Zade: Maybe I can hold it up for the technical round, but yeah, in terms of the AI solutions that we are developing, is it some sort of agentic workflows that we are doing, or some RAGs, or chatbots, or any such sort of things?
72 00:13:36.520 ⇒ 00:13:38.719 Nevan Zade: If you, if you know the mix around that.
73 00:13:38.980 ⇒ 00:13:40.869 Kaela Gallagher: Yeah,
74 00:13:41.230 ⇒ 00:13:59.520 Kaela Gallagher: I know that we have a mixture there as well. I know there’s clients that we’ve done, like, chat features with, there’s other clients that we’ve done more, like, agentic solutions with, so again, I would say it’s kind of dependent on the client. When we’re hiring, we’re really looking for people that have, like.
75 00:13:59.650 ⇒ 00:14:06.179 Kaela Gallagher: A wide range of experience, but can also just, like, learn and adapt really quickly to whatever our clients might need.
76 00:14:07.290 ⇒ 00:14:13.999 Nevan Zade: Sounds good, sounds good, sounds good, yeah. Yeah, I think the current scenario is everything is around AI, and you get to deliver.
77 00:14:14.400 ⇒ 00:14:15.410 Kaela Gallagher: Another one.
78 00:14:15.410 ⇒ 00:14:17.510 Nevan Zade: Yeah, definitely. I do, I do get that.
79 00:14:17.680 ⇒ 00:14:26.349 Nevan Zade: Sounds good, sounds good. And in terms of the team structure itself, like, how does the team look like? Like you said, it is such a small team that’s working with Reinforge.
80 00:14:26.660 ⇒ 00:14:38.349 Kaela Gallagher: Yeah, so as a company, we’re about 25 people right now, and like I mentioned, like, most of us are split into either, like, strategy, data, or AI, service lines.
81 00:14:38.550 ⇒ 00:14:42.939 Kaela Gallagher: So yeah, overall, like, a total of 25-ish.
82 00:14:43.860 ⇒ 00:14:47.100 Nevan Zade: Sounds good, sounds good. I have gone over the website, and it doesn’t feel like…
83 00:14:47.310 ⇒ 00:14:50.400 Nevan Zade: 25 people are there, it looks pretty mature.
84 00:14:50.400 ⇒ 00:14:54.269 Kaela Gallagher: I know, I thought the same thing, too, when I was joining.
85 00:14:55.000 ⇒ 00:15:04.820 Nevan Zade: Sounds cool, sounds cool. Yeah, I think it’s a good one, and given that I have experience with the consultancies, and most of the teams I have worked with are ranging from 25 to 50, so…
86 00:15:04.950 ⇒ 00:15:19.480 Nevan Zade: it’s good to see that pace, because I really enjoy that agile type of development that we do, and we do this incremental, in an incremental way. I think that I can hold on for later, to confirm that, like, what’s the development style that we follow.
87 00:15:19.740 ⇒ 00:15:27.440 Nevan Zade: But yeah, sounds cool, sounds cool. I think, yeah, that’s pretty much I had to ask you. Anything you want to confirm from my side?
88 00:15:27.950 ⇒ 00:15:38.200 Kaela Gallagher: I don’t think I have anything else for you. What I’ll do is I’ll go ahead and, just review your profile with… with my team, and then I’ll get back to you on whether or not we’ll move forward.
89 00:15:39.120 ⇒ 00:15:39.940 Nevan Zade: Sounds good, sounds cool.
90 00:15:39.940 ⇒ 00:15:43.410 Kaela Gallagher: Okay, awesome. Nevin, thanks so much for your time, I appreciate it.
91 00:15:44.400 ⇒ 00:15:45.690 Nevan Zade: Definitely. Have a good one.
92 00:15:45.900 ⇒ 00:15:47.519 Kaela Gallagher: Have a good one. Bye.