Meeting Title: Brainforge AI Engineer Interview Date: 2026-03-04 Meeting participants: read.ai meeting notes, Samuel Roberts, Daniel Ángel
WEBVTT
1 00:03:02.490 ⇒ 00:03:03.940 Daniel Ángel: Hi, Samuel.
2 00:03:04.250 ⇒ 00:03:05.110 Samuel Roberts: Hello!
3 00:03:06.140 ⇒ 00:03:07.400 Daniel Ángel: Hello, can you hear me?
4 00:03:09.880 ⇒ 00:03:12.740 Daniel Ángel: I can hear you, give me a second.
5 00:03:19.190 ⇒ 00:03:21.180 Daniel Ángel: My cross…
6 00:03:26.050 ⇒ 00:03:27.280 Daniel Ángel: Can you hear me?
7 00:03:27.280 ⇒ 00:03:28.270 Samuel Roberts: Yes, I can.
8 00:03:29.050 ⇒ 00:03:30.849 Daniel Ángel: Okay, okay, now he’s fine.
9 00:03:31.190 ⇒ 00:03:32.670 Samuel Roberts: Great. How are you?
10 00:03:33.260 ⇒ 00:03:35.180 Daniel Ángel: Hey, great. How about you?
11 00:03:35.600 ⇒ 00:03:36.620 Samuel Roberts: Good, good.
12 00:03:36.820 ⇒ 00:03:43.629 Samuel Roberts: Yeah, so, brief intro, I guess. I’m Sam Roberts, I’m the AI Tech Lead here at Brainforge.
13 00:03:43.920 ⇒ 00:03:53.380 Samuel Roberts: So I think, you know, we need to start maybe with a brief intro from you, and then I got some questions. I’ll leave some time for you to ask some questions, and then,
14 00:03:53.500 ⇒ 00:03:57.500 Samuel Roberts: Yeah, that’ll be the interview, so… Hey, sure. If you could just give me a brief intro on yourself?
15 00:03:58.270 ⇒ 00:04:10.630 Daniel Ángel: Sure, sure. Well, I… I am Danielle Angel. I have around 12 years of experience as a full-stack developer. I work in both eyes, in the front, in the back.
16 00:04:10.650 ⇒ 00:04:29.150 Daniel Ángel: But in the last years, I have been working more in the back, for example, creating LLX integrations, apps like creating agents with tools like… or frameworks like LandChain, LandGraph.
17 00:04:29.380 ⇒ 00:04:41.379 Daniel Ángel: crew AI, using models like… Cloud, cloud, from OpenAI, Gemini.
18 00:04:41.780 ⇒ 00:04:50.279 Daniel Ángel: Languages that I haven’t used, are ROS, NO, Express.
19 00:04:50.610 ⇒ 00:05:07.640 Daniel Ángel: In the last 3 years, I have been using a lot Python for the integration with the LLMs, OpenAI libraries. I also have experience with blockchain. In one moment, I wore a lot with blockchain, creating smart contracts with Solidity.
20 00:05:07.640 ⇒ 00:05:12.140 Daniel Ángel: So, I… I… I got good knowledge about that.
21 00:05:12.280 ⇒ 00:05:21.330 Daniel Ángel: But right now, I am working more on enhancing my skill with LLMs, integration, agents, and things like that.
22 00:05:21.780 ⇒ 00:05:26.659 Daniel Ángel: My native language is Spanish, but I try to do my best with the English.
23 00:05:26.930 ⇒ 00:05:34.440 Samuel Roberts: Yeah, yeah, you’re doing well. Can you tell me about, an LLM-based feature that you’ve shipped to production, and what problem it solved?
24 00:05:35.660 ⇒ 00:05:44.599 Daniel Ángel: Well, in one moment, I received an error from the connection with the LLMs.
25 00:05:44.670 ⇒ 00:05:46.020 Samuel Roberts: Okay.
26 00:05:46.480 ⇒ 00:05:59.199 Daniel Ángel: I fit it by… because I was using LandChain and LandGraph, so you can track what is happening in the background when the LLN is calling.
27 00:05:59.200 ⇒ 00:06:13.389 Daniel Ángel: So, I thought, was a problem with the API keys, so one of them was that the API key, for example, for Brave Search.
28 00:06:13.660 ⇒ 00:06:30.949 Daniel Ángel: It’s a service or tool that you can use for searching internet with your LLMs, with your agents, and I received high error because I didn’t pay for that. So I saw it paying a plan.
29 00:06:31.350 ⇒ 00:06:45.830 Daniel Ángel: And then it was working fine. Also, in the prong of the agent, I tried to clarify if something happened in the background while trying to connect with the LLMs and the tools.
30 00:06:46.090 ⇒ 00:06:59.389 Daniel Ángel: in the problem I put that gives me a good answer about what is the problem. So the LLN also gives me the errors, for example, 400 errors.
31 00:06:59.620 ⇒ 00:07:04.300 Daniel Ángel: So with that, I detect what was the bug.
32 00:07:04.520 ⇒ 00:07:05.360 Daniel Ángel: I don’t think…
33 00:07:06.950 ⇒ 00:07:13.959 Samuel Roberts: Great. What part of the, the AI stack have you spent the most time building on versus just experimenting with?
34 00:07:13.960 ⇒ 00:07:14.510 Daniel Ángel: 5th?
35 00:07:15.670 ⇒ 00:07:17.060 Daniel Ángel: Can you repeat that again, please?
36 00:07:17.060 ⇒ 00:07:27.600 Samuel Roberts: Yeah, so what part of the overall stack have you spent the most time actually building, like, production apps, versus just experimenting with on, you know, just to learn?
37 00:07:27.600 ⇒ 00:07:34.870 Daniel Ángel: One production app that I did was, AI agent assistant.
38 00:07:35.090 ⇒ 00:07:41.159 Daniel Ángel: It was for… For use, for my… for myself.
39 00:07:41.380 ⇒ 00:08:00.639 Daniel Ángel: For example, if something… if somebody can want to know what I do, what is my background, my experience, they use this agent to ask anything about myself. I deploy it in human phase, in a space.
40 00:08:00.720 ⇒ 00:08:05.029 Daniel Ángel: It’s working fine, so,
41 00:08:05.430 ⇒ 00:08:21.100 Daniel Ángel: It answers about me, you know, questions like, like, like this. Also, I put a tool to send push notification to my phone with an app that sends push notifications.
42 00:08:21.200 ⇒ 00:08:36.959 Daniel Ángel: If they ask about something that the agent didn’t know about… don’t know about me, it will send me a push notification that a new question that the agent doesn’t have.
43 00:08:36.960 ⇒ 00:08:47.400 Daniel Ángel: I can put that answer for the agent when somebody come again with… with a question.
44 00:08:48.100 ⇒ 00:09:01.999 Daniel Ángel: For testings and to get more knowledge about LLNs and agents and things like that, I have been creating a,
45 00:09:02.160 ⇒ 00:09:12.170 Daniel Ángel: like, apps, LLMs, or AI apps, with Crew AI, for example, is a framework to create agents.
46 00:09:12.220 ⇒ 00:09:26.139 Daniel Ángel: Lunch and LandGraph, Microsoft AutoGen also is another framework. But in… I prefer to use more Python libraries.
47 00:09:26.390 ⇒ 00:09:32.580 Daniel Ángel: For example, the OpenAI library and things like that, I feel more comfortable with that.
48 00:09:33.660 ⇒ 00:09:35.550 Daniel Ángel: Waters? Yeah.
49 00:09:35.930 ⇒ 00:09:49.670 Samuel Roberts: Okay, yeah, so kind of shifting a little bit. So we work with, you know, different clients that are hearing all about this AI stuff, and they have all kinds of ideas, but they don’t necessarily… they’re non-technical. They don’t necessarily know
50 00:09:50.110 ⇒ 00:09:58.709 Samuel Roberts: what actually is possible. How do you go about explaining the limitations of LLMs and other tools to non-technical stakeholders?
51 00:10:00.560 ⇒ 00:10:04.590 Daniel Ángel: Well, they, they, they, when they don’t have,
52 00:10:05.220 ⇒ 00:10:10.929 Daniel Ángel: experience with the technical part, although they don’t know about that.
53 00:10:11.050 ⇒ 00:10:17.249 Daniel Ángel: We can try to explain how the workflow works.
54 00:10:17.430 ⇒ 00:10:27.230 Daniel Ángel: how to use it, maybe same documentation about that, to read more how to use it.
55 00:10:27.350 ⇒ 00:10:34.649 Daniel Ángel: create a doc where it’s very explained everything, the process, how to use the tools, the agents.
56 00:10:35.030 ⇒ 00:10:50.230 Daniel Ángel: With that, you can ensure the clients or the people that doesn’t have, technical knowledge use the agents, the tools, the apps that we developed.
57 00:10:50.700 ⇒ 00:10:59.649 Daniel Ángel: Yes, no, right now, you can find many AI tools that can help you with any question.
58 00:11:00.380 ⇒ 00:11:01.240 Samuel Roberts: Cool.
59 00:11:01.720 ⇒ 00:11:05.620 Samuel Roberts: So yeah, let’s talk about some newer tools. When there’s something…
60 00:11:06.010 ⇒ 00:11:16.729 Samuel Roberts: new, or a new trend, or framework, or model? Was there been anything that you’ve been excited about, but decided not to adopt for some… some reason?
61 00:11:17.980 ⇒ 00:11:27.099 Daniel Ángel: Yes, well, I am a person that always try to learn the latest tech things.
62 00:11:29.720 ⇒ 00:11:40.030 Daniel Ángel: when I saw… I… when I see something that is new and is helping a lot in the tech industry, I try to learn about that.
63 00:11:40.280 ⇒ 00:11:45.830 Daniel Ángel: And try to integrate the company one, something like that.
64 00:11:45.940 ⇒ 00:11:55.489 Daniel Ángel: Or study it by myself to get the knowledge and use in one moment in the… in the future.
65 00:11:55.710 ⇒ 00:12:13.010 Daniel Ángel: normally, my… I try to study about that by reading the doc. I, like, watch YouTube videos, tutorials. I buy a lot Udemy course. It’s one of my preferences.
66 00:12:13.200 ⇒ 00:12:21.209 Daniel Ángel: prefer platform to learn. I think they have… Udemy have good, teachers about…
67 00:12:21.450 ⇒ 00:12:26.330 Daniel Ángel: things like that. I got many certifications from this platform.
68 00:12:26.700 ⇒ 00:12:28.290 Daniel Ángel: Yes.
69 00:12:29.280 ⇒ 00:12:29.890 Samuel Roberts: Okay.
70 00:12:29.990 ⇒ 00:12:32.240 Samuel Roberts: Let’s see…
71 00:12:37.100 ⇒ 00:12:45.339 Samuel Roberts: Okay, so actually, we’re getting kind of close to the halfway point here, so I guess I want to open it up, make sure I give time for you to ask any questions. So,
72 00:12:45.800 ⇒ 00:12:49.460 Samuel Roberts: Yeah, if you have any questions about Brainforge, about the role, I’m happy to.
73 00:12:49.460 ⇒ 00:12:57.259 Daniel Ángel: Well, I would like to know what are you doing in your company, Brain Fork?
74 00:12:57.490 ⇒ 00:13:09.120 Daniel Ángel: this role is about, AI engineer, or something like that. Yeah. Are you using Python? Framework for the agents?
75 00:13:09.120 ⇒ 00:13:13.059 Samuel Roberts: Yeah, we use… we use some Python, we use a lot of TypeScript as well.
76 00:13:13.270 ⇒ 00:13:13.700 Daniel Ángel: No.
77 00:13:13.700 ⇒ 00:13:21.970 Samuel Roberts: Because we build, you know, UIs and front-ends, and so we kind of just try to stick with that when it’s a full-stack kind of thing, but there is some Python.
78 00:13:22.140 ⇒ 00:13:31.249 Samuel Roberts: We’ve been using, mastra is the framework we’ve been using for TypeScript.
79 00:13:31.460 ⇒ 00:13:34.499 Samuel Roberts: Even though we’ve used some Langchain when we use Python.
80 00:13:36.920 ⇒ 00:13:38.870 Samuel Roberts: Yeah, we use, we use,
81 00:13:39.200 ⇒ 00:13:41.410 Samuel Roberts: A bunch of different models we play with.
82 00:13:41.540 ⇒ 00:13:47.520 Samuel Roberts: So we have a bunch on Azure for OpenAI, we have a bunch, we’re setting up some Gemini ones now.
83 00:13:49.000 ⇒ 00:13:51.339 Samuel Roberts: Yeah. And we do different work for different clients.
84 00:13:51.340 ⇒ 00:13:57.420 Daniel Ángel: building, like, agents, workflows, things like that.
85 00:13:57.420 ⇒ 00:14:16.549 Samuel Roberts: Yeah, so, it’s interesting. We have, kind of internal work that we do to help the rest of the company, and then we have client work that we do for external clients. And so, depending on what kind of clients we have at any given time, we’re building different kinds of agents, different kinds of automations, so some were, like, a RAG-assisted.
86 00:14:17.010 ⇒ 00:14:18.919 Samuel Roberts: Excuse me, chat agent.
87 00:14:20.440 ⇒ 00:14:32.859 Samuel Roberts: Another one was more of an automation for a client that kind of already had a process, and they were just copying and pasting things into Claude over and over, and so we kind of automated that for them, so that was less of an agent, more of a flow.
88 00:14:32.970 ⇒ 00:14:44.399 Samuel Roberts: We’ve built some UIs, some platforms, we’re building one internally that we have that ingests all our meetings, and you can search over transcripts, and, you know, if you missed a meeting, you can catch up.
89 00:14:44.580 ⇒ 00:14:49.199 Samuel Roberts: Yeah, we’re experimenting more with actual just, you know, agentic flows,
90 00:14:49.560 ⇒ 00:14:51.270 Samuel Roberts: You know, we built a case study.
91 00:14:51.270 ⇒ 00:14:54.500 Daniel Ángel: Go ahead. Some exciting.
92 00:14:54.590 ⇒ 00:14:57.180 Samuel Roberts: Yes, yes, we’re doing a lot of cool stuff, yeah.
93 00:14:57.180 ⇒ 00:14:58.600 Daniel Ángel: Yeah, right.
94 00:14:58.710 ⇒ 00:15:05.810 Daniel Ángel: You’re arguing the time zone, same tan mi? For example, I am in Venezuela.
95 00:15:06.100 ⇒ 00:15:10.359 Samuel Roberts: Yeah, I’m in East Coast time in the US, so I think it’s minus 5.
96 00:15:11.150 ⇒ 00:15:18.970 Samuel Roberts: Yeah, I think so. There’s some people in the West Coast, and some people in the Central, so we kind of cover the U.S. times, pretty much.
97 00:15:19.160 ⇒ 00:15:19.920 Daniel Ángel: Okay.
98 00:15:20.080 ⇒ 00:15:25.890 Daniel Ángel: And this is the first, interview, then how many steps is needed?
99 00:15:25.890 ⇒ 00:15:39.219 Samuel Roberts: Yeah, so, the next step would be another more role-based technical interview, and then after that, I believe there’s a tech challenge, and then a panel interview about that, with three of us from the team.
100 00:15:39.400 ⇒ 00:15:52.180 Samuel Roberts: We like to move pretty quickly, so you usually hear back pretty quickly, then just trying to schedule is usually what takes the most time, and sort of, you know, fitting into people’s schedules, but, yeah, we don’t like to drag it out too much, so…
101 00:15:52.180 ⇒ 00:15:56.579 Daniel Ángel: Okay, great. And the budget for this role, do you know what is that?
102 00:15:56.580 ⇒ 00:16:01.529 Samuel Roberts: I do not know offhand.
103 00:16:01.740 ⇒ 00:16:03.370 Daniel Ángel: But I’m sure we can…
104 00:16:04.490 ⇒ 00:16:07.149 Samuel Roberts: find that out, or I can check the…
105 00:16:07.150 ⇒ 00:16:09.639 Daniel Ángel: No vase que yo.
106 00:16:10.160 ⇒ 00:16:14.770 Samuel Roberts: Yeah, I don’t see… Yeah, I don’t think I have that right now, I’m sorry, but…
107 00:16:15.420 ⇒ 00:16:16.530 Daniel Ángel: No problem.
108 00:16:17.380 ⇒ 00:16:19.660 Samuel Roberts: Alright, any other questions then, or…
109 00:16:19.960 ⇒ 00:16:36.009 Daniel Ángel: No, I will wait for the next steps. I like what are you doing in the background with LLMs, agents. It’s something that is very exciting to continue learning and integrating in production environments.
110 00:16:36.010 ⇒ 00:16:42.450 Daniel Ángel: So, I will wait if I continue with the ne- for the next steps.
111 00:16:42.960 ⇒ 00:16:44.150 Samuel Roberts: Great. Sounds good.
112 00:16:44.150 ⇒ 00:16:44.750 Daniel Ángel: Okay.
113 00:16:44.810 ⇒ 00:16:46.379 Samuel Roberts: Alright, thank you for the time.
114 00:16:46.590 ⇒ 00:16:47.919 Daniel Ángel: Thank you. Bye.
115 00:16:47.920 ⇒ 00:16:48.510 Samuel Roberts: Bye-bye.