Meeting Title: Awaish <> Vishnu Tech Screen Date: 2025-06-12 Meeting participants: Awaish Kumar, Vishnu, Fireflies.ai Notetaker Awaish
WEBVTT
1 00:00:16.280 ⇒ 00:00:17.230 Vishnu: Hello!
2 00:00:27.120 ⇒ 00:00:27.980 Vishnu: Hello!
3 00:00:29.040 ⇒ 00:00:30.659 Vishnu: Am I audible?
4 00:00:30.660 ⇒ 00:00:31.270 Awaish Kumar: Yep.
5 00:00:32.620 ⇒ 00:00:35.020 Awaish Kumar: Hello! How are you?
6 00:00:36.377 ⇒ 00:00:37.920 Vishnu: I’m good. How about you, sir.
7 00:00:39.670 ⇒ 00:00:41.940 Awaish Kumar: I’m good as well, so.
8 00:00:43.570 ⇒ 00:00:46.738 Vishnu: Yes, thanks for getting on a call. And
9 00:00:47.550 ⇒ 00:00:49.461 Vishnu: you know, doing this with me.
10 00:00:49.910 ⇒ 00:00:50.595 Vishnu: So
11 00:00:52.110 ⇒ 00:01:02.416 Vishnu: yeah, I was talking to so just I’ll give you a little bit 2 seconds about me 2 min about me, and then yeah,
12 00:01:02.880 ⇒ 00:01:04.370 Vishnu: we’ll take it from there.
13 00:01:05.930 ⇒ 00:01:13.710 Awaish Kumar: Yeah, like, thank you for joining. And yeah, I would like to just introduce little like little bit about what is
14 00:01:14.360 ⇒ 00:01:24.289 Awaish Kumar: like, what is the agenda of this meeting like? Why, what we are going to talk in this meeting, and something about Brentforge. And and then we can start from there.
15 00:01:25.110 ⇒ 00:01:25.680 Vishnu: Okay. Okay.
16 00:01:25.680 ⇒ 00:01:27.280 Awaish Kumar: So yeah.
17 00:01:28.110 ⇒ 00:01:38.610 Awaish Kumar: like it, the brain for basically, it’s an AI power data analytics company. We are providing Eia and data services to different clients.
18 00:01:39.181 ⇒ 00:01:42.269 Awaish Kumar: and we have a team of like
19 00:01:42.600 ⇒ 00:01:47.109 Awaish Kumar: they are in total now, like maybe 12 to 15 people, which.
20 00:01:47.640 ⇒ 00:01:55.509 Awaish Kumar: Which basically spend all across departments like data, analyst analytics, engineering AI
21 00:01:55.670 ⇒ 00:02:01.309 Awaish Kumar: team and operations and marketing sales. So
22 00:02:02.030 ⇒ 00:02:09.700 Awaish Kumar: like, that’s mainly the kind of services we provide. It’s like, go into the client’s data
23 00:02:09.850 ⇒ 00:02:30.100 Awaish Kumar: and then try to figure out, build a data warehouse and how and and build the dashboards, and like, provide the like. The insights like how they can extract insight from the data. And apart from that, like, there are some of the AI services which are like built using the
24 00:02:30.390 ⇒ 00:02:32.990 Awaish Kumar: like Lms or.
25 00:02:34.450 ⇒ 00:02:51.650 Awaish Kumar: Rag or something to to build some agents and automate the processes we like. We have some clients work, plan work. And then. So we have some internal projects where we basically automate, the like most of the things which we do
26 00:02:52.279 ⇒ 00:03:03.860 Awaish Kumar: like, we, we haven’t kind of enabled everyone in our in our company even for for like non technical teams we have enabled everybody to use
27 00:03:04.404 ⇒ 00:03:15.019 Awaish Kumar: Ai to help speed up their processes to speed up their day to day work. We have built the automation projects, and then also
28 00:03:15.662 ⇒ 00:03:20.900 Awaish Kumar: like something like you can say, like summarizing things and out of
29 00:03:22.440 ⇒ 00:03:34.946 Awaish Kumar: llm work. So yeah, that’s what what the brain forge is doing. And then in the in this meeting, like, it’s mainly about understanding you and
30 00:03:35.520 ⇒ 00:03:43.680 Awaish Kumar: understanding about your previous experiences. What kind of projects you have worked on, and what kind of skill set have you possess.
31 00:03:44.630 ⇒ 00:03:48.129 Awaish Kumar: And yeah, you can start ahead with your introduction.
32 00:03:49.470 ⇒ 00:04:01.979 Vishnu: Yes, okay, so yeah, I’m I’m basically a new fresh grad from data science from George Washington University before that.
33 00:04:02.790 ⇒ 00:04:04.380 Vishnu: Hmm. So.
34 00:04:04.680 ⇒ 00:04:21.050 Vishnu: firstly, yeah, thank you for taking time to like, speak with me. And you know, explain what brain Forge does, and see if like my skills align with what brain Forge is doing, and see if like I can fit into your company. And so
35 00:04:21.339 ⇒ 00:04:43.178 Vishnu: I I did. My bachelor’s in mechanical engineering, and then so I studied and call in this college called Savior in India. And so back in my early days I’ve been always somebody who’s curious about asking the basic foundational questions of like, hey? Why does this work like this? Or why is a certain part failing and
36 00:04:43.710 ⇒ 00:04:55.658 Vishnu: asking the question and trying to pursue, you know available resources to push in that direction, to see if we are able to like. Answer that in a you know data driven
37 00:04:56.270 ⇒ 00:05:12.059 Vishnu: or like scientifically backed manner. At least, that’s what like. Because my both my parents are in academic research, and that’s something that they have instilled in me. And so so that’s about me doing my undergrad. And so after that.
38 00:05:12.800 ⇒ 00:05:32.380 Vishnu: I did fall in love with like storytelling. And like that’s also the time when you know I was part of the Student Council so I used to run like this club for quizzes. So I used to do whole like trivia quizzes. And I started writing blogs around those I started writing on this website called Medium.
39 00:05:32.610 ⇒ 00:05:48.729 Vishnu: So back in 2020, I started writing on medium. And I used to write, I mean, I still write till this day. But so I write mainly about you know, history culture. And then I also blog about tech here and there.
40 00:05:49.260 ⇒ 00:05:57.170 Vishnu: So when I say tech, I I blog about like, you know python programming SQL, and
41 00:05:58.564 ⇒ 00:06:09.955 Vishnu: something along you know a lot of things along those lines. And so when I used to blog on medium I, I did see that a lot of my articles were getting a lot of traction. And
42 00:06:10.730 ⇒ 00:06:14.520 Vishnu: I did end up being one of the top 1,000 writers on that platform.
43 00:06:14.790 ⇒ 00:06:15.400 Vishnu: So
44 00:06:16.190 ⇒ 00:06:28.935 Vishnu: yeah. So medium did send me like a a recognition for that. And then I did get awarded a couple of times for being one of the top engaging writers on their platform. And
45 00:06:29.870 ⇒ 00:06:58.149 Vishnu: so this is where you know. The, this is where my 1st job, experience comes in. Because, why, when I used to write articles on medium a founder from so a founder from like this Yc. Back Startup. His name was Akash, so he reached out to me and he said. He likes my you know tech blogs he likes the way my other blogs are written. And he asked me, Hey, would you be interested in coming and doing marketing for our team.
46 00:06:58.500 ⇒ 00:06:59.350 Vishnu: Oh.
47 00:07:00.020 ⇒ 00:07:19.498 Vishnu: so so that’s how I got my 1st break into the industry. So this this company was called Jovian. So at Jovian I, I wore like 3 hats. So I used to maintain, like, I used to maintain that data analytics code base like they had. So Jovian was like this,
48 00:07:20.287 ⇒ 00:07:38.160 Vishnu: course, boot camp. Course. Selling platform. At least, that’s what they we. That was our edge like this thing. The agenda of company was to sell as many boot camps as possible, and so I I joined in as like a data science software generalist.
49 00:07:38.270 ⇒ 00:07:41.194 Vishnu: But then my forte was marketing. So
50 00:07:42.580 ⇒ 00:08:08.568 Vishnu: so I during the day, I’m working with the software team. I’m working with the marketing team. And I also I was, I single handedly used to like grade submissions that students from all across the world would like submit. So when I say submissions, I mean, like project submissions quiz submissions so Jovian had this very comprehensive quiz, and so after each
51 00:08:08.970 ⇒ 00:08:17.764 Vishnu: you know lesson they would have like this quiz, which would last 45 min, and it was an open ended quiz. It was not Mcq’s. So
52 00:08:18.980 ⇒ 00:08:42.290 Vishnu: students could record or or voice notes. Students could type down. So I would like grade those so in the one year, one and one year. 2 months that I spent at Jovian I like ended up validating over 130 plus you know, user generated notebooks. This includes projects, too. So projects along the fields of AI
53 00:08:42.809 ⇒ 00:08:55.273 Vishnu: okay, back in the day, we didn’t have the term AI, we we just called it. Machine learning projects and deep learning projects. Ai is all the reads as of now, right? So
54 00:08:56.440 ⇒ 00:09:06.638 Vishnu: And when I say when I look back. And I I look at the marketing aspect. How did I contribute? So when I joined Jovian, Jovian was like,
55 00:09:07.200 ⇒ 00:09:24.243 Vishnu: Jovian was at ground 0 like they didn’t have any followers on Twitter, on Linkedin on Youtube. So I helped you know, make up this whole marketing campaign. And by the time I left Jovian we had over
56 00:09:24.960 ⇒ 00:09:28.050 Vishnu: close to 12,000 followers on Linkedin.
57 00:09:28.210 ⇒ 00:09:32.509 Vishnu: We had around 1010 K followers on Youtube.
58 00:09:32.690 ⇒ 00:09:52.230 Vishnu: And so the Youtube channel was just catching traction. And again, before I I joined Jovian. The marketing team was completely like they. They would not use data to make decisions like. So when I joined Jovian, like, the 1st thing I did was, you know, scrape Youtube data.
59 00:09:52.230 ⇒ 00:10:04.988 Vishnu: scrape challenge channels that were in our domain. And then, try and compare like, hey, what videos in this domain are performing? What videos are getting traction. So engagement? So
60 00:10:05.740 ⇒ 00:10:06.620 Vishnu: like,
61 00:10:07.910 ⇒ 00:10:17.430 Vishnu: okay, there is this one video with a million views it has a thousand comments. So what did they do? That? We are not doing? So
62 00:10:18.030 ⇒ 00:10:36.189 Vishnu: yeah. So something along those lines. And when I joined the marketing team had again, the marketing team had no clue about data using data and making decisions. So I kind of tried to help, at least so I didn’t do do this all by myself. When I look back.
63 00:10:36.190 ⇒ 00:10:50.023 Vishnu: There was also the founders who were guiding me, so each time I would have an idea I would 1st pitch it to them, and they would in turn validate it, and they would tell, Hey, can you make these changes? Can you tweak it here and there? And
64 00:10:50.510 ⇒ 00:10:55.430 Vishnu: so that was my time at Jovi, and so by the time I left Jovian like
65 00:10:56.720 ⇒ 00:11:04.609 Vishnu: we had boosted like altogether, we had boosted like Jovian’s growth, like the sales were up 42
66 00:11:04.820 ⇒ 00:11:08.265 Vishnu: to 45%. From what I remember, like the
67 00:11:08.980 ⇒ 00:11:14.692 Vishnu: the followers, count had gone up like the platform was just catching traction. So
68 00:11:15.430 ⇒ 00:11:23.171 Vishnu: again that, you know, really enforced reinforced the fact that hey? When you use data like in in very small
69 00:11:24.203 ⇒ 00:11:28.899 Awaish Kumar: How many like you have full time, experience working there, or.
70 00:11:30.632 ⇒ 00:11:45.609 Vishnu: Yeah. So my my role at Jovian was an internship. But then, since this was like this was around the time of the Covid pandemic. So I used to work full time there like I had flown into Bangalore, and the.
71 00:11:45.610 ⇒ 00:11:46.810 Awaish Kumar: What was your time like?
72 00:11:46.920 ⇒ 00:11:50.839 Awaish Kumar: You work for 2 months or no, I won’t.
73 00:11:50.840 ⇒ 00:11:51.220 Vishnu: Like.
74 00:11:51.220 ⇒ 00:11:52.069 Awaish Kumar: What was the duration?
75 00:11:52.070 ⇒ 00:11:53.849 Vishnu: One year, 2 months.
76 00:11:54.440 ⇒ 00:11:55.120 Awaish Kumar: Okay?
77 00:11:55.680 ⇒ 00:11:57.160 Awaish Kumar: And that was like.
78 00:11:58.050 ⇒ 00:12:06.839 Awaish Kumar: okay, like, I want to understand, like you mentioned internship is that like, is, was it a job? A full time job, or was it an internship.
79 00:12:09.055 ⇒ 00:12:25.124 Vishnu: So I I joined as an intern because I I was when I joined Jovian. I was also a full time student in college, so my college wouldn’t allow me to like skip classes and be at office 20 or like 7 days a week. But then
80 00:12:26.070 ⇒ 00:12:37.379 Vishnu: since the log Covid Lockdown had hit just then, and I had gone to Bangalore. I got to like stay in their office and like work with them. And so we had a team of 7 people.
81 00:12:37.590 ⇒ 00:12:42.340 Vishnu: And so it it was just us a team of 7 people, and we were running.
82 00:12:42.340 ⇒ 00:12:44.369 Awaish Kumar: Okay. So you join them as a
83 00:12:44.570 ⇒ 00:12:47.389 Awaish Kumar: as a like, what was your role.
84 00:12:47.390 ⇒ 00:12:50.270 Vishnu: A data software generalist role.
85 00:12:50.760 ⇒ 00:12:55.179 Vishnu: So I joined as a data intern. But then I also used to do a lot of like software.
86 00:12:55.180 ⇒ 00:12:58.880 Awaish Kumar: So mainly what what was like, what I understood from your
87 00:12:59.090 ⇒ 00:13:03.980 Awaish Kumar: what you say it is that like you were
88 00:13:04.340 ⇒ 00:13:08.600 Awaish Kumar: helping them with marketing stuff and
89 00:13:09.930 ⇒ 00:13:12.999 Awaish Kumar: but and your help also, like kind of
90 00:13:13.110 ⇒ 00:13:19.650 Awaish Kumar: reviewing some notebooks for the students which was which were coming. So I want like.
91 00:13:20.360 ⇒ 00:13:27.509 Awaish Kumar: like when you say, like, you are helping with marketing and stuff like, what, exactly, was your role in that.
92 00:13:28.340 ⇒ 00:13:56.848 Vishnu: Yes, so my role was so I would again. So when I would work with the marketing team, the the marketing team did not have much idea about analytics. They did not. Or even if they had, they were not implementing. It is what I would say. So what I did as you know, Hiery, there was. My 1st task was to make Ada notebooks like I said I would. I scraped like data from Youtube. And
93 00:13:57.270 ⇒ 00:14:16.794 Vishnu: so I, I remember, like we had this notebook where we I compared like close to 1,000 channels in that particular domain, and these 1,000 channels were across the board, like some international, some Indian, some you know, some from Europe. So different different channels along those lines, and
94 00:14:17.410 ⇒ 00:14:20.182 Vishnu: I made an Edn notebook, and I did a presentation.
95 00:14:20.460 ⇒ 00:14:20.780 Awaish Kumar: The.
96 00:14:20.780 ⇒ 00:14:21.640 Vishnu: Marketing team.
97 00:14:22.020 ⇒ 00:14:30.249 Awaish Kumar: You scrap the Youtube videos and their titles, how people are engaging with with that.
98 00:14:30.250 ⇒ 00:14:31.040 Vishnu: Yes, yes.
99 00:14:31.040 ⇒ 00:14:36.060 Awaish Kumar: Videos and the content of the video. Or, Oh.
100 00:14:36.060 ⇒ 00:14:46.849 Vishnu: So when I say, content of the video like it was like hashtags, title and based of the Youtube description, the video description.
101 00:14:47.360 ⇒ 00:14:48.010 Awaish Kumar: Okay.
102 00:14:48.010 ⇒ 00:14:53.020 Vishnu: Yes, and you know, pulling out keywords from there and seeing. Okay, what is the.
103 00:14:53.020 ⇒ 00:14:57.399 Awaish Kumar: Oh, how did you scrap this Youtube videos?
104 00:14:58.250 ⇒ 00:15:04.569 Vishnu: So there were like couple of ways how I did it! 1st I used the Api, then I used selenium.
105 00:15:04.860 ⇒ 00:15:13.479 Vishnu: so the Api would often get blocked. So then, I just use selenium to like, go and automatically scrape the web pages for me.
106 00:15:13.650 ⇒ 00:15:17.719 Vishnu: And yeah, selenium is basically what worked out.
107 00:15:17.890 ⇒ 00:15:22.608 Vishnu: And apart from that. I also like
108 00:15:23.020 ⇒ 00:15:23.909 Awaish Kumar: But like how you.
109 00:15:23.910 ⇒ 00:15:24.410 Vishnu: I don’t.
110 00:15:24.410 ⇒ 00:15:28.169 Awaish Kumar: You know, just like you open the selenium, search for
111 00:15:28.420 ⇒ 00:15:31.970 Awaish Kumar: keywords, and then scrub the data, or like how.
112 00:15:32.680 ⇒ 00:15:41.176 Vishnu: Yes, so there was like this bunch of keywords that we were targeting. So we had like we had an excel sheet where
113 00:15:41.880 ⇒ 00:16:01.387 Vishnu: I had put in some keywords, and also the marketing team, like we had 2 other people who would come and look at it like they would, they would update the keywords every day. And so the selenium script would run for like 7 7 days at least. That’s what I remember. Like we we let it run for 7 days, and we let it like,
114 00:16:01.900 ⇒ 00:16:14.580 Vishnu: you know, scrape data. And this is this was back in the day in during the lockdown. So a lot of you know, people were like trying to sell courses online. A lot of people were doing you know.
115 00:16:16.190 ⇒ 00:16:32.349 Vishnu: Hackathon like web hackathons, data thons like they would continuously stream. So which is why, rather than just running for one day the script, we let it run across 7 days, and then once a week, we would again update it. So that kind of stuff.
116 00:16:32.842 ⇒ 00:16:42.827 Vishnu: Yeah. So that was what we did. That was what I did with the marketing team. And now coming with the software team like so this was a a.
117 00:16:43.160 ⇒ 00:16:44.530 Awaish Kumar: I like. What?
118 00:16:44.630 ⇒ 00:16:49.320 Awaish Kumar: So with selenium? How like? What was your infrastructure?
119 00:16:49.550 ⇒ 00:16:53.020 Awaish Kumar: How are you running those that person like? Where were you running
120 00:16:53.220 ⇒ 00:16:57.869 Awaish Kumar: like, I want to understand this full like the technical things about that project.
121 00:16:59.910 ⇒ 00:17:00.460 Vishnu: So
122 00:17:01.600 ⇒ 00:17:10.599 Vishnu: so selenium, we basically like so I ran it on the computer, like on the laptop that we had. And so
123 00:17:10.900 ⇒ 00:17:13.680 Vishnu: when you say which infrastructure.
124 00:17:16.460 ⇒ 00:17:21.760 Awaish Kumar: Where did you run like? I have my computer, I can develop on a script which basically.
125 00:17:22.230 ⇒ 00:17:26.299 Awaish Kumar: Opens a browse like uses selenium to open a browser to
126 00:17:26.750 ⇒ 00:17:29.339 Awaish Kumar: open the Youtube. There’s something but
127 00:17:29.700 ⇒ 00:17:32.990 Awaish Kumar: like, How do you deploy it like
128 00:17:33.100 ⇒ 00:17:36.590 Awaish Kumar: how it runs for product like the one you’re running on your own own computer.
129 00:17:37.620 ⇒ 00:17:44.500 Vishnu: Yeah. So we were basically running this on my own so on the work computer that we had. I used to run it on that, and then.
130 00:17:44.500 ⇒ 00:17:45.690 Awaish Kumar: Over your like.
131 00:17:46.080 ⇒ 00:17:51.489 Awaish Kumar: You didn’t have Github, or you didn’t push it to some servers where it is running, or.
132 00:17:53.190 ⇒ 00:18:05.311 Vishnu: No, we we didn’t. So it wasn’t live on Github. And yeah, so basically, I’ll I’ll walk you through the whole workflow. So so the the script was written on on our id, and then
133 00:18:05.750 ⇒ 00:18:18.909 Vishnu: at at certain time, like the script would start on its own, and it would like you said it would open the browser and the selenium would like go to Youtube. And then it would like call the
134 00:18:19.751 ⇒ 00:18:21.740 Vishnu: you know keywords from the excel.
135 00:18:22.140 ⇒ 00:18:22.600 Vishnu: We have.
136 00:18:22.600 ⇒ 00:18:28.909 Awaish Kumar: That I that I understand like how script works I was. I’m I’m interested in the in the knowing, like
137 00:18:29.380 ⇒ 00:18:34.350 Awaish Kumar: how like I I cannot run something on my computer, right? It’s not
138 00:18:34.550 ⇒ 00:18:41.285 Awaish Kumar: safe. It’s it’s not possible I have to do other things. My laptop might not be on every time.
139 00:18:42.790 ⇒ 00:18:45.980 Awaish Kumar: So we we normally deploy it somewhere.
140 00:18:45.980 ⇒ 00:18:55.599 Vishnu: Yes, I get what you’re saying. So the ideal ways of like putting it on a Vm. Or like putting it on an instance, and like you know so that
141 00:18:55.910 ⇒ 00:19:02.100 Vishnu: at a certain point in time, like when it has to run, it has to run right. So the whole. You know,
142 00:19:02.920 ⇒ 00:19:14.907 Vishnu: having github actions, and like having the pipeline ready. But in this case, because we were like scratchy, we were like a small company, at least, like I was also student. I was learning. So
143 00:19:15.790 ⇒ 00:19:19.140 Vishnu: we had that computer 24 by 7 connected. And it was on.
144 00:19:19.140 ⇒ 00:19:19.600 Awaish Kumar: For me.
145 00:19:19.600 ⇒ 00:19:20.330 Vishnu: And.
146 00:19:21.200 ⇒ 00:19:27.140 Awaish Kumar: Okay, so like, after that, do you have any experience after after that job like.
147 00:19:28.850 ⇒ 00:19:32.059 Vishnu: Yes, yes, so so this was my time at Jovi.
148 00:19:32.060 ⇒ 00:19:37.709 Awaish Kumar: What is your recent like? Recently? You you are you. You have graduated like you have done master’s.
149 00:19:38.730 ⇒ 00:19:40.449 Vishnu: and before that.
150 00:19:41.070 ⇒ 00:19:46.569 Awaish Kumar: And before the master, like how many years of experience you have in total.
151 00:19:47.600 ⇒ 00:20:12.390 Vishnu: So most of my experience. So it’s like internship experiences I haven’t really gone and like worked at a full time company. So my! My time at Jovian was like one year 2 months, and then I had like a startup, which was like a so I’ll tell you about my startup in India. This was called Stellar Nursery, and what stellar nursery was like a vertical farming startup.
152 00:20:12.550 ⇒ 00:20:30.027 Vishnu: So again, stellar nursery worked for a year. So a quick 2 min about stellar nursery, what stellar nursery was so it was kind of this wild wild ride where I was like the intra lead. And basically the guy who did everything from, you know, idea pitching and like
153 00:20:30.650 ⇒ 00:20:41.759 Vishnu: generating funds to like run this company. So what we were trying to do is like help farmers, you know, detect crop diseases using computer vision. And
154 00:20:42.310 ⇒ 00:20:56.933 Vishnu: my role, there was again, you know, analyzing 17. You know, there were like 1718 companies in India which did vertical farming. So I was like, Hey, how can we build tech for like automating vertical farming. And
155 00:20:57.680 ⇒ 00:21:09.002 Vishnu: so I use like again, SQL stats and like scraping company data from, we had a bunch of news websites that would publish about vertical farming. And
156 00:21:09.780 ⇒ 00:21:15.034 Vishnu: here, my, I here my role was more about pitching to the University
157 00:21:15.910 ⇒ 00:21:29.287 Vishnu: Innovation Center, and like getting funds. We had, like 3 students who used to build the Ml. Ml. Related stuff, you know, the whole compute convolutional neural network pipeline. So
158 00:21:29.670 ⇒ 00:21:35.520 Awaish Kumar: You. You wanted to have this disease detection using vision.
159 00:21:38.000 ⇒ 00:21:41.619 Awaish Kumar: Ml, only for vertical farming.
160 00:21:43.530 ⇒ 00:22:08.069 Vishnu: Yes, when I say disease, detection, there’s a little bit more to it. So when plants grow, and when you’re growing plants vertically in, like, you know, in a vertical farming setup. So it it goes through different growth stages. So the whole point about doing vertical farming was to reduce or renew human intervention, so the bot has to be able to identify.
161 00:22:08.380 ⇒ 00:22:16.269 Vishnu: At what growth state is this plant like, hey? Is the plant a seedling? Is the plant at a certain mid range. So
162 00:22:16.390 ⇒ 00:22:29.170 Vishnu: let’s say you’re growing lettuce right? So when you harvest lettuce at 2 weeks the the leaves are not completely formed. So the when a user bites into the leaf it is not crunchy enough.
163 00:22:29.370 ⇒ 00:22:53.060 Vishnu: and so you let it, you know, grow for 2 and a half weeks. Now the leaf is really crunchy. So when you put it in a sandwich like the user or the person who’s having that, you will feel the difference, and because it’s grown vertically, there is no pest. There is no, you know, other agent like variables that are like spoiling the plant.
164 00:22:53.210 ⇒ 00:22:53.980 Vishnu: Oh.
165 00:22:54.330 ⇒ 00:22:54.980 Awaish Kumar: No.
166 00:22:54.980 ⇒ 00:22:55.740 Awaish Kumar: Okay.
167 00:22:55.740 ⇒ 00:22:56.220 Vishnu: So.
168 00:22:56.855 ⇒ 00:22:58.770 Awaish Kumar: Like, understand?
169 00:23:00.061 ⇒ 00:23:07.890 Awaish Kumar: Yeah, like, I, I would like to get more into the the work you have done for the data
170 00:23:08.130 ⇒ 00:23:17.730 Awaish Kumar: as were discussing before. Like, so what I understand that, like you haven’t deployed anything so far like.
171 00:23:18.670 ⇒ 00:23:30.409 Awaish Kumar: To somewhere, and then in the university, like recently in your master’s, like, what kind of projects you work with, which are related to data, or Gen. AI, or right.
172 00:23:31.970 ⇒ 00:23:40.680 Awaish Kumar: And what kind of tools like you have used. Python Asql, any cloud platforms or anything there.
173 00:23:42.230 ⇒ 00:23:43.580 Vishnu: Okay, so
174 00:23:44.140 ⇒ 00:23:53.399 Vishnu: So now, my 2 years of masters, I I’ve like worked on. I’ve worked on various projects. And so first, st to begin with,
175 00:23:55.122 ⇒ 00:23:57.479 Vishnu: let me just share a link.
176 00:23:57.480 ⇒ 00:23:59.030 Awaish Kumar: No like. Just
177 00:23:59.680 ⇒ 00:24:08.019 Awaish Kumar: just let me know about the project which you are most proud of, and which is related to data.
178 00:24:09.110 ⇒ 00:24:15.539 Awaish Kumar: Engineering data analysis, or something related to data, science or something.
179 00:24:16.880 ⇒ 00:24:20.759 Vishnu: Okay, nomadly, is my project 1 min?
180 00:24:20.950 ⇒ 00:24:21.770 Vishnu: Well.
181 00:24:23.530 ⇒ 00:24:30.765 Vishnu: so I had built this travel recommendation engine, which was called nomadly, let me just pull up the link for it.
182 00:24:34.999 ⇒ 00:24:53.309 Vishnu: okay, yeah. Oh, sorry. The instance is currently off. Because, the university, like the Aws accounts that we were using. So okay, I’ll tell you quickly about my project nomadly. So so nomadly is like what I imagined in my head was me and my teammate. We were 2 people.
183 00:24:53.310 ⇒ 00:25:04.768 Vishnu: So we were like, Hey, can we build a a you know, a travel recommendation engine kind of like Chat Gpt, you know, which is like a full blown data pipeline that starts from Kafka ingestion. And
184 00:25:05.370 ⇒ 00:25:23.420 Vishnu: it runs like spark streaming with like validation like Z ordering compactation, and like. Then it stores it in date. It stores it in delta, and then we again validate it, using this thing called Great Expectations. And so
185 00:25:23.890 ⇒ 00:25:35.214 Vishnu: we then use that data to like fine tune this llama, one B model, and the llama one B model was deployed on the Ec. 2 instance, and
186 00:25:35.850 ⇒ 00:25:47.280 Vishnu: so there were. We had like this feedback loop, where every week the llama model would like fine tune again. And for this fine tuning we used race rate train.
187 00:25:47.630 ⇒ 00:25:48.285 Vishnu: And
188 00:25:50.670 ⇒ 00:26:10.309 Vishnu: we serve the model using reserve. And why did we use retrain and reserve, because I had like read a blog on medium again. That retrain and reserve are like the industry norms, and you know, retraining and deploying chat. Gpt. Kind of models.
189 00:26:11.660 ⇒ 00:26:12.620 Vishnu: So
190 00:26:13.160 ⇒ 00:26:19.855 Vishnu: for the front end we built it, using next year tailwind Css, and there were 2 aspects to our
191 00:26:21.410 ⇒ 00:26:23.180 Vishnu: 1 min. Where is this?
192 00:26:24.220 ⇒ 00:26:26.649 Vishnu: Is there a way I can? Yeah, 1 min.
193 00:26:32.880 ⇒ 00:26:35.310 Vishnu: I’m just trying to share the link with you.
194 00:26:36.910 ⇒ 00:26:37.690 Vishnu: Almost.
195 00:26:38.860 ⇒ 00:26:46.379 Awaish Kumar: Okay, where? But I will. Yeah, you can like you. You had a system where you people can check. And then, like.
196 00:27:08.640 ⇒ 00:27:22.140 Vishnu: So. You know, what with nomadly we did not want it to give, like, you know, generic. You know, top 10 things to do in DC kind of list. But we wanted something, you know, personalized context of it, and conversational
197 00:27:22.509 ⇒ 00:27:28.740 Vishnu: something that is a mix between a travel agent and a friend who knows your wife. Kind of. At least that was the idea.
198 00:27:28.740 ⇒ 00:27:29.210 Awaish Kumar: Sorry.
199 00:27:29.210 ⇒ 00:27:30.289 Vishnu: Started off with.
200 00:27:31.890 ⇒ 00:27:37.489 Awaish Kumar: The link like here. You I cannot chat right. It’s it’s just like some options, and
201 00:27:37.920 ⇒ 00:27:39.779 Awaish Kumar: it generates some recommendation.
202 00:27:40.550 ⇒ 00:27:47.549 Vishnu: There are 2 aspects there is like the hard coded options, and then there is like the chat bot oops.
203 00:27:48.520 ⇒ 00:27:51.369 Vishnu: If you click AI itinerary creator.
204 00:27:51.910 ⇒ 00:27:52.640 Awaish Kumar: Okay.
205 00:27:57.680 ⇒ 00:28:03.849 Vishnu: So yeah, the AI itinerary creator basically is the Chatbot feature of it.
206 00:28:06.220 ⇒ 00:28:11.369 Awaish Kumar: But when I type something and then generate itinerary. Okay? So I still.
207 00:28:12.300 ⇒ 00:28:16.550 Awaish Kumar: okay, it’s like it takes the input as in a human format.
208 00:28:16.680 ⇒ 00:28:21.180 Awaish Kumar: and then goes calls to some Llm.
209 00:28:21.340 ⇒ 00:28:21.980 Awaish Kumar: Returns.
210 00:28:21.980 ⇒ 00:28:25.600 Vishnu: Yeah. It calls to the yes, yes, that’s right.
211 00:28:25.820 ⇒ 00:28:29.089 Awaish Kumar: Okay, I cannot continue chatting with this.
212 00:28:29.270 ⇒ 00:28:34.540 Awaish Kumar: Okay, after my 1st one. Now, I want to say something else I want to modify.
213 00:28:34.750 ⇒ 00:28:36.900 Awaish Kumar: It’s not a chat box right.
214 00:28:38.598 ⇒ 00:28:58.621 Vishnu: Yes, so it does not. Have context in the sense. So again, when I began the project, the idea was so that we have a AI agent or a a travel recommendation engine that has context about all your chats. But then, for us to have all your chats told, and like, serve that again. It was costing a lot. And I also
215 00:28:59.840 ⇒ 00:29:00.690 Vishnu: oh.
216 00:29:00.800 ⇒ 00:29:13.529 Vishnu: like. So this instance currently like the Ec 2 instance where this project, like the whole data pipeline was deployed is currently not working because I graduated college and like they cut access to it.
217 00:29:13.840 ⇒ 00:29:17.758 Vishnu: And so. But yes, what you said is right.
218 00:29:18.400 ⇒ 00:29:32.009 Vishnu: it does not. It does not store your past chat and talk on top of that. It just generates one chat, and if you ask it another question, it will again generate a new chat, but which is not continuous to the past one.
219 00:29:33.453 ⇒ 00:29:35.369 Vishnu: So, okay, yeah, thanks for.
220 00:29:35.370 ⇒ 00:29:41.679 Awaish Kumar: You understand, like you can, but but like you know how to retain the context, and.
221 00:29:42.330 ⇒ 00:29:43.650 Vishnu: Yes. Yes.
222 00:29:44.370 ⇒ 00:29:44.750 Vishnu: Yeah.
223 00:29:44.750 ⇒ 00:29:47.970 Awaish Kumar: Okay. So how would you do that? If if we have to build something
224 00:29:48.180 ⇒ 00:29:50.299 Awaish Kumar: where we want to retain the
225 00:29:50.440 ⇒ 00:29:54.029 Awaish Kumar: like in the in the session like whatever we have talked.
226 00:29:54.560 ⇒ 00:29:57.029 Awaish Kumar: we just want to keep that.
227 00:29:58.363 ⇒ 00:30:06.236 Vishnu: So when I think about retaining context, firstly, you should have we should store you know the
228 00:30:06.750 ⇒ 00:30:32.318 Vishnu: conversation or session id for each of those user chats and that session id, I would store it in a database like you know, Mongodb or Redis, for example. And then, when I ask a new question, it should retrieve the previous message, so that now all the previous messages build Context window in that for that, Llm. And you know it concatenates that
229 00:30:32.940 ⇒ 00:30:39.850 Vishnu: last chat to the next chat in each of the Llm. Calls right. And again, when
230 00:30:39.990 ⇒ 00:30:46.592 Vishnu: when you add up a lot of chats there there has to be proper memory management. So that
231 00:30:47.190 ⇒ 00:30:58.979 Vishnu: if if let’s say, somebody has 10 chats, 10 concurrent chat consecutive chats then I am imagining that the context limit of that Llm. Will be breached.
232 00:30:59.130 ⇒ 00:31:08.130 Vishnu: So there has to be a certain logic where, like I am summarizing and doing proper memory management, so that tokens are not exceeded
233 00:31:09.360 ⇒ 00:31:09.900 Vishnu: so.
234 00:31:09.900 ⇒ 00:31:11.070 Awaish Kumar: And you can just.
235 00:31:12.930 ⇒ 00:31:13.470 Vishnu: Oh!
236 00:31:13.470 ⇒ 00:31:17.150 Awaish Kumar: Like you can utilize like filtering to filter out.
237 00:31:18.310 ⇒ 00:31:18.764 Vishnu: Yes.
238 00:31:19.220 ⇒ 00:31:25.200 Awaish Kumar: The messages which are closer to your new message. Only use that in the countries.
239 00:31:26.450 ⇒ 00:31:37.682 Vishnu: Yes, yeah, yes, correct. I would use that, or I I would also like to use, you know, using pulling out certain keywords and summarizing it. And
240 00:31:39.300 ⇒ 00:31:50.829 Vishnu: yeah, in my head, I am imagining multiple approaches that I would take to this to do proper memory management and keeping the context in the limited token size
241 00:31:51.880 ⇒ 00:31:54.575 Vishnu: again. So
242 00:31:57.108 ⇒ 00:32:11.151 Vishnu: I would use like some summary comp compression to like summarize the older context for longer chats. And you know, on the front end I would send the session, be with user messages. So back end can fetch and assemble that in the right
243 00:32:11.740 ⇒ 00:32:13.280 Vishnu: back. Then.
244 00:32:13.280 ⇒ 00:32:18.579 Awaish Kumar: About AI. So what what different talents you have worked with.
245 00:32:21.855 ⇒ 00:32:22.600 Vishnu: So
246 00:32:22.940 ⇒ 00:32:30.749 Vishnu: I had this class for for deep learning. So we we did build rag rag based Llms from scratch.
247 00:32:30.890 ⇒ 00:32:38.756 Vishnu: But while I did build rag based llm from scratch, I would say, like,
248 00:32:40.330 ⇒ 00:32:43.098 Vishnu: that wasn’t really my forte, because
249 00:32:45.518 ⇒ 00:32:49.743 Vishnu: simply because, you know. We were.
250 00:32:51.240 ⇒ 00:32:52.586 Vishnu: you know we were.
251 00:32:53.520 ⇒ 00:33:00.062 Vishnu: we were stuck by the amount of compute that we had, the amount of time that we had, and
252 00:33:00.580 ⇒ 00:33:04.040 Awaish Kumar: Like what kind of different elements you have used, like
253 00:33:04.340 ⇒ 00:33:09.850 Awaish Kumar: the already trained versions like Llama to make the
254 00:33:10.240 ⇒ 00:33:18.230 Awaish Kumar: some some from open AI, some from Google. So what can? What, what different you have like that you have used.
255 00:33:19.980 ⇒ 00:33:29.569 Vishnu: Oh, okay, so, okay, what different Llms I have used. So I have basically used all the Llms that are in market. And like I I still use it till date. So
256 00:33:30.410 ⇒ 00:33:38.849 Vishnu: I’ve I’ve used llama one B I’ve actually like use one B and 8 B both. I’ve tried to fine tune for different use cases.
257 00:33:38.940 ⇒ 00:34:01.498 Vishnu: And I’ve used open AI’s Api. I’ve used deep seek Api for so I I didn’t tell you about this. So for the last 2 years I have been working with a doctor in my university, and he he has this whole database of patient data he has like research papers. So he wanted to build again an Llm. For his application. Where
258 00:34:02.080 ⇒ 00:34:17.750 Vishnu: he could like. When when he when a new student comes. They can like, you know. Look at Screw, you know, screw screen through all the data that is there and like train on it. So for that particular application I used open AI to like
259 00:34:18.850 ⇒ 00:34:23.645 Vishnu: Oh, firstly, there were 2 aspects about this. There were 2 so
260 00:34:24.130 ⇒ 00:34:26.920 Vishnu: no, I can actually share you the link for this one, too.
261 00:34:30.489 ⇒ 00:34:31.630 Vishnu: Just curious.
262 00:34:31.639 ⇒ 00:34:39.589 Awaish Kumar: And apart from the AI, what like in terms of data, analyt analytics.
263 00:34:39.819 ⇒ 00:34:41.259 Vishnu: What, what.
264 00:34:41.609 ⇒ 00:34:45.409 Awaish Kumar: Like, how do you rate yourself in python and SQL.
265 00:34:45.759 ⇒ 00:34:46.829 Awaish Kumar: Out of 10.
266 00:34:49.325 ⇒ 00:34:58.250 Vishnu: So my python skills I would like rate myself 8 and SQL, I would rate myself again 7.5 7.
267 00:34:59.340 ⇒ 00:34:59.720 Awaish Kumar: Okay.
268 00:34:59.720 ⇒ 00:35:02.790 Vishnu: And yes, so.
269 00:35:03.870 ⇒ 00:35:04.510 Awaish Kumar: So? How.
270 00:35:04.510 ⇒ 00:35:05.000 Vishnu: This.
271 00:35:05.423 ⇒ 00:35:09.500 Awaish Kumar: Hope I can handle the memory management.
272 00:35:12.850 ⇒ 00:35:13.220 Vishnu: Okay,
273 00:35:14.820 ⇒ 00:35:19.999 Vishnu: So when you say memory management like.
274 00:35:21.350 ⇒ 00:35:24.740 Vishnu: how like how python handles memory management?
275 00:35:25.020 ⇒ 00:35:25.660 Vishnu: Oh.
276 00:35:25.660 ⇒ 00:35:30.629 Awaish Kumar: Like in the python. You define variables, you define lists, you define functions.
277 00:35:30.770 ⇒ 00:35:39.609 Awaish Kumar: All of this needs memory. So how python handles it like memory in the back end.
278 00:35:41.560 ⇒ 00:35:44.973 Vishnu: Okay? So from what I remember, like,
279 00:35:45.530 ⇒ 00:35:50.299 Vishnu: python has like this automatic memory management framework, right?
280 00:35:50.440 ⇒ 00:35:55.271 Vishnu: And so so it employs like this.
281 00:35:55.830 ⇒ 00:35:58.409 Vishnu: in memory garbage collector, which, like
282 00:35:59.510 ⇒ 00:36:11.379 Vishnu: manages, allocates memory, deallocate, allocates memory for like objects, and then it uses a ref count like for object tracking reference. And
283 00:36:11.980 ⇒ 00:36:12.595 Vishnu: so
284 00:36:14.750 ⇒ 00:36:28.517 Vishnu: so there are certain locations on the memory where, like if I define a variable, it, it puts that location. It allocates that variable, that location on the memory. And then it keeps updating that location.
285 00:36:30.190 ⇒ 00:36:31.210 Vishnu: oh.
286 00:36:31.340 ⇒ 00:36:42.870 Vishnu: it keeps updating that location. So if if I update a value on the variable, that it does not create a new variable there, it updates that location on the memory on the RAM,
287 00:36:43.080 ⇒ 00:36:44.872 Vishnu: and so
288 00:36:46.820 ⇒ 00:36:50.609 Vishnu: It uses cyclic like references for
289 00:36:51.450 ⇒ 00:36:53.989 Vishnu: cycling through referencing each way.
290 00:36:53.990 ⇒ 00:36:55.459 Awaish Kumar: But what about the scopes?
291 00:36:55.610 ⇒ 00:36:57.089 Awaish Kumar: Do you know the scope?
292 00:36:57.580 ⇒ 00:37:07.500 Awaish Kumar: Local scope, global scope like impacts in Python? So when you define a variable, how.
293 00:37:08.980 ⇒ 00:37:12.200 Awaish Kumar: Python handles the scope of that variable.
294 00:37:14.270 ⇒ 00:37:39.556 Vishnu: Okay? Yeah. So in python, like, variables are accessible using like, yes, like, you said global and like local scopes. So when you say local scopes. If a variable is defined within a function and is accessible only in that function, then it is like a local function like a local scope, right? And a global is when I am defining it.
295 00:37:40.420 ⇒ 00:37:50.039 Vishnu: I’m defining at the module level where it is accessible at every part in that particular in that particular module. So it’s outside the function.
296 00:37:50.440 ⇒ 00:37:55.439 Vishnu: And but so if I define a if I define a list.
297 00:37:55.850 ⇒ 00:38:00.240 Awaish Kumar: In a function like in the main. In the main.
298 00:38:00.530 ⇒ 00:38:04.070 Awaish Kumar: I define a list, and I pass this list into the function.
299 00:38:05.590 ⇒ 00:38:08.300 Awaish Kumar: And then inside of function you are
300 00:38:08.470 ⇒ 00:38:15.290 Awaish Kumar: trying to update. Add some more values to that list, and then the function ends.
301 00:38:16.600 ⇒ 00:38:18.469 Awaish Kumar: and we are now back in the main.
302 00:38:19.050 ⇒ 00:38:21.590 Awaish Kumar: and we we define a functionality there.
303 00:38:22.290 ⇒ 00:38:29.560 Awaish Kumar: so will the state of that array change, or it will be the.
304 00:38:29.560 ⇒ 00:38:38.082 Vishnu: Okay. So yes. If if you reassign the list which is inside the function, then it won’t affect the original list, whereas
305 00:38:38.470 ⇒ 00:38:41.789 Awaish Kumar: So, for example, we say, you know what I mean.
306 00:38:42.508 ⇒ 00:38:49.430 Awaish Kumar: Block, I defined a list named Eddie and I
307 00:38:49.840 ⇒ 00:38:55.440 Awaish Kumar: assigned 2 2 Va values in 2 integer values like one and 2
308 00:38:56.426 ⇒ 00:39:06.589 Awaish Kumar: inside of that list. So in the main block, I only have an array with 2 items. Now, I pass it to a function
309 00:39:06.760 ⇒ 00:39:11.840 Awaish Kumar: which basically has a parameter, and in that in the function I just
310 00:39:12.090 ⇒ 00:39:17.660 Awaish Kumar: use that parameter and the dot append function of the list, and I append more values.
311 00:39:18.170 ⇒ 00:39:24.739 Awaish Kumar: When that function ends, we back. We are back in the main block men code block and then
312 00:39:24.950 ⇒ 00:39:27.770 Awaish Kumar: and then I say, print, array.
313 00:39:28.010 ⇒ 00:39:39.949 Awaish Kumar: So what will be the output? Will it be only one like the one and 2 which I defined before that function, or will it also print the values appended inside of that function?
314 00:39:42.050 ⇒ 00:39:44.040 Vishnu: Okay? So
315 00:39:45.556 ⇒ 00:40:10.119 Vishnu: so when, again, if I’m understanding modifications inside that function do persist outside, right? So if I am making modification in the list which is inside a main. Then it does. When I call it outside that function, it it does. It does bring back like, pass that list out. So when I pass a list to a function, the function it receives
316 00:40:10.840 ⇒ 00:40:14.854 Vishnu: is a reference to that whole list. And so any
317 00:40:16.660 ⇒ 00:40:28.049 Vishnu: any imputations like addition, like like, yes, if I append, if I like update or remove elements inside that function, it will be reflected in that list in the main scope.
318 00:40:29.310 ⇒ 00:40:30.040 Awaish Kumar: Okay.
319 00:40:32.253 ⇒ 00:40:32.976 Vishnu: So
320 00:40:34.200 ⇒ 00:40:42.200 Awaish Kumar: Yeah of the guard after the okay? And what are the contacts? Managers.
321 00:40:42.200 ⇒ 00:40:48.640 Vishnu: Original list it. It won’t change the original list. It is just the local reference inside that function that changes right.
322 00:40:51.240 ⇒ 00:40:53.549 Awaish Kumar: It will change the original list right?
323 00:40:53.780 ⇒ 00:41:00.580 Awaish Kumar: Because when you pass a list in a parameter, it actually referencing the original object.
324 00:41:00.700 ⇒ 00:41:02.669 Awaish Kumar: it is not creating a new object.
325 00:41:04.720 ⇒ 00:41:05.540 Vishnu: Oh.
326 00:41:05.920 ⇒ 00:41:13.509 Vishnu: yes. Sorry. What I meant is like, yes, yeah. Yeah. So a a new list of that parameter inside the function.
327 00:41:14.170 ⇒ 00:41:14.560 Vishnu: like.
328 00:41:15.980 ⇒ 00:41:20.020 Awaish Kumar: I just said, I just passed this as a parameter. I never said, I will create a new list.
329 00:41:20.590 ⇒ 00:41:21.610 Vishnu: Yes. Yeah.
330 00:41:22.930 ⇒ 00:41:27.580 Awaish Kumar: But okay, and what do you? What do you know about context managers?
331 00:41:29.700 ⇒ 00:41:33.334 Vishnu: Oh, so context, managers
332 00:41:35.550 ⇒ 00:41:43.007 Vishnu: so in pi, like python context, managers are mainly like used for like file operations, right? And
333 00:41:44.670 ⇒ 00:41:47.769 Vishnu: so when I use.
334 00:41:48.870 ⇒ 00:41:59.920 Vishnu: Yeah. So when I would use like enter or like for setting up and like exit for cleaning up methods, I would use it for like files for resource management, basically like about.
335 00:41:59.920 ⇒ 00:42:01.439 Awaish Kumar: No, no! How we use it.
336 00:42:01.670 ⇒ 00:42:06.890 Awaish Kumar: What you will write in Python. What, what? Exactly, what keywords do we write.
337 00:42:07.420 ⇒ 00:42:18.290 Vishnu: I would write context, context, library, Const context lib, which is like, with the with statements about like, what file operations I want to do for let’s say.
338 00:42:19.250 ⇒ 00:42:23.210 Awaish Kumar: Can you repeat, for example, if I prefer.
339 00:42:23.210 ⇒ 00:42:27.110 Vishnu: With open file text as as.
340 00:42:28.910 ⇒ 00:42:33.520 Awaish Kumar: Okay, you will use it with with keyword and
341 00:42:34.180 ⇒ 00:42:37.449 Awaish Kumar: and inside you will have some things, and you to resor.
342 00:42:37.450 ⇒ 00:42:39.449 Awaish Kumar: So we’ll and is the resource. We will have
343 00:42:39.660 ⇒ 00:42:43.860 Awaish Kumar: context managers. So context managers are basically to use
344 00:42:44.320 ⇒ 00:42:51.249 Awaish Kumar: to efficiently utilize the resources. And yeah, and the clean up after we are done with the.
345 00:42:51.250 ⇒ 00:42:52.080 Vishnu: You know. Yes.
346 00:42:52.080 ⇒ 00:42:52.660 Awaish Kumar: Processing.
347 00:42:53.640 ⇒ 00:43:04.610 Awaish Kumar: Okay, and in terms of SQL, like how like cool laptop
348 00:43:07.010 ⇒ 00:43:13.870 Awaish Kumar: like apart, like apart from SQL, like what
349 00:43:15.313 ⇒ 00:43:21.170 Awaish Kumar: tools for the warehouses? Have you used to like
350 00:43:22.100 ⇒ 00:43:27.680 Awaish Kumar: to basically retain the data? Or we transform the data.
351 00:43:30.400 ⇒ 00:43:32.004 Vishnu: Okay so
352 00:43:34.130 ⇒ 00:43:38.040 Vishnu: So you’re you’re asking me what other tools I have used? Right? So.
353 00:43:38.040 ⇒ 00:43:42.769 Awaish Kumar: I, I’m saying, like what databases or data warehouses have you used?
354 00:43:43.050 ⇒ 00:43:48.139 Awaish Kumar: And obviously, for all the data warehouses and databases like
355 00:43:48.680 ⇒ 00:43:53.050 Awaish Kumar: you have, do you have to use SQL. To to query them?
356 00:43:53.210 ⇒ 00:44:04.510 Awaish Kumar: And so one question is, what kind of what different data warehouses and databases have you used? And have you used things like SQL. Mesh or dbt.
357 00:44:06.180 ⇒ 00:44:07.680 Vishnu: Yes, so
358 00:44:08.230 ⇒ 00:44:29.906 Vishnu: again, when when I talk about my experience, like I I’m I’m more towards the data science like presenting dashboards, tableau dashboards, power bi doing exploratory data analysis that you know, basically ab testing statistical analysis and like presenting it to the stakeholders and helping them push their decision making in the right direction. But
359 00:44:30.280 ⇒ 00:44:40.385 Vishnu: so when you talk about like databases I have experience on postgres, I have experience using SQL lite, I have experience
360 00:44:41.472 ⇒ 00:44:47.377 Vishnu: non relational so I have mongodb experience. I’ve used readers. And
361 00:44:49.383 ⇒ 00:44:54.529 Vishnu: when when you talk about the other tools I have experience in hadoop spark.
362 00:44:54.690 ⇒ 00:45:01.860 Vishnu: And yeah. So for my other project, I’ve used Kafka for the whole.
363 00:45:01.860 ⇒ 00:45:03.909 Vishnu: No, no, just the databases.
364 00:45:06.250 ⇒ 00:45:11.180 Vishnu: Okay? So yeah. So databases. Oh.
365 00:45:11.640 ⇒ 00:45:13.279 Awaish Kumar: Or data, warehouses.
366 00:45:15.130 ⇒ 00:45:22.619 Vishnu: So yeah, so databases, this is basically what I’ve used. I’ve also, so Delta Lake is one data warehouse snowflake
367 00:45:22.960 ⇒ 00:45:25.080 Awaish Kumar: Data lake is a data leak.
368 00:45:25.900 ⇒ 00:45:26.890 Vishnu: Delta, Leak.
369 00:45:27.270 ⇒ 00:45:29.889 Awaish Kumar: Yeah. Delta lake is a data lake.
370 00:45:29.890 ⇒ 00:45:30.480 Vishnu: Yes.
371 00:45:30.480 ⇒ 00:45:34.150 Awaish Kumar: Basically is data like with asset properties.
372 00:45:34.750 ⇒ 00:45:38.050 Vishnu: Yes, yes, and.
373 00:45:38.050 ⇒ 00:45:41.140 Awaish Kumar: And I don’t know.
374 00:45:41.150 ⇒ 00:45:43.199 Vishnu: Yes, I have. I have.
375 00:45:44.230 ⇒ 00:45:49.310 Awaish Kumar: So like, okay, I have last few questions. I just want to understand.
376 00:45:49.920 ⇒ 00:46:01.399 Awaish Kumar: If you join Brain Forge as an intern. What? Exactly you you are looking to work on. You want to work on AI projects. Do you want to work on.
377 00:46:01.550 ⇒ 00:46:02.460 Awaish Kumar: Tom?
378 00:46:03.570 ⇒ 00:46:10.600 Awaish Kumar: Some like dashboard building stuff or
379 00:46:10.770 ⇒ 00:46:17.420 Awaish Kumar: more into the analytics engineering, like creating transformations and processing data
380 00:46:17.900 ⇒ 00:46:21.580 Awaish Kumar: or data engineering part like ingestion and the
381 00:46:23.630 ⇒ 00:46:27.249 Awaish Kumar: writing automate automation pipelines or things like that.
382 00:46:28.010 ⇒ 00:46:46.399 Vishnu: Okay, so based off my past experiences like I, I still have a couple of experiences that I haven’t talked to you about but then, based off those experiences based off me, working on like this research project, I would prefer building like Ml. Pipelines again. I had talked to utham and
383 00:46:48.020 ⇒ 00:47:17.759 Vishnu: pious about the same, because my whole point about what would Vishnu contribute? Vishnu would contribute with the analytics part Vishnu. Can you know automating analytics Llm based solutions, you know, productionizing data science models. That is my forte. But then, what I want to learn from projects that are there at data, you know, at Brainforge, I I also said Utham, that I want like a mentor who can guide me with learning data engineering.
384 00:47:17.990 ⇒ 00:47:20.493 Vishnu: because, I haven’t,
385 00:47:21.200 ⇒ 00:47:29.974 Vishnu: you know, gone in depth into data engineering. All all this while I have been working on the surface level with analytics, dashboarding. And
386 00:47:30.760 ⇒ 00:47:38.901 Vishnu: you know, using statistical analysis to again guide decision making, solving like real business questions. So
387 00:47:39.880 ⇒ 00:47:50.589 Vishnu: yes, what I would work love to work on is like end to end. AI, you know, AI ml workflow like it. It includes data, pipelines, model training, deployment.
388 00:47:51.400 ⇒ 00:47:58.070 Awaish Kumar: So like, we have related projects. And we have then analytical related projects.
389 00:47:58.466 ⇒ 00:48:24.909 Awaish Kumar: And something like we, we mentioned about data engineering, like data building data pipelines. Like we for example, we want to get data from, we have similar things as you mentioned about helping marketing team. So we have similar pro projects where we want to. We have linkedin data of our posts. We want to analyze them. So we have data somewhere. We want to bring that in in our data warehouse. And then we want to analyze our
390 00:48:25.100 ⇒ 00:48:32.289 Awaish Kumar: how our engagement is on the Linkedin or Youtube, or different on different platforms. And then.
391 00:48:32.991 ⇒ 00:48:35.198 Awaish Kumar: how can we actually
392 00:48:35.960 ⇒ 00:48:38.019 Awaish Kumar: improve it? Or things like that?
393 00:48:39.266 ⇒ 00:48:45.329 Awaish Kumar: So this is kind of end to end pipeline. So num like the data engineering is like.
394 00:48:45.580 ⇒ 00:48:58.165 Awaish Kumar: let’s bring the data in from different platforms. How we are going to bring that in like writing scripts, deploying it somewhere or using 3rd party tools, whatever it is, you know, whatever it takes to get there and then,
395 00:48:58.590 ⇒ 00:49:05.059 Awaish Kumar: keep keep like, get the data somewhere in our warehouse in our storage, in our fillet file somewhere
396 00:49:05.677 ⇒ 00:49:23.119 Awaish Kumar: and then build it transformations on top of it. Writing SQL. In the Dbt. And and after we have transformed tables, we will. We will be building the dashboards finally, maybe using tableau or power bi or some open source solution like Apache percent.
397 00:49:23.260 ⇒ 00:49:24.219 Awaish Kumar: a bunch of superset.
398 00:49:24.220 ⇒ 00:49:24.710 Vishnu: Okay.
399 00:49:24.710 ⇒ 00:49:27.240 Awaish Kumar: And like things like that. So
400 00:49:28.450 ⇒ 00:49:48.060 Awaish Kumar: so we like, these are the parts we are working on in working on right now. And it’s, it’s just one. It’s. It’s just one example of a project. But it’s like things like that. We are on. You know, data engineering. We are on data analysis, we are on our on AI Projects building Llms building, automating.
401 00:49:48.170 ⇒ 00:50:00.990 Awaish Kumar: how we for example, build, how we do project management or things like that. But then, what you mentioned about ml, like Ml. Is is right now. It’s kind of
402 00:50:03.090 ⇒ 00:50:06.570 Awaish Kumar: kind of kind of shadowed behind AI and Lms.
403 00:50:06.950 ⇒ 00:50:19.920 Awaish Kumar: Right now. Companies are not not more. Not talking about. Ml, I have worked on Ml. Myself. I have built the models, deployed them in production. I have like deployed real time models batch models.
404 00:50:20.090 ⇒ 00:50:23.839 Awaish Kumar: But the problem is that right now the the companies are
405 00:50:24.543 ⇒ 00:50:36.620 Awaish Kumar: but more looking for like, okay, we’ll have some Llm like, build something AI Llm, and do some things which which is not like.
406 00:50:37.100 ⇒ 00:50:43.639 Awaish Kumar: which is not a solution for everything. But that’s what the buzz is in the market right now.
407 00:50:43.750 ⇒ 00:50:55.800 Awaish Kumar: And a lot of projects we are getting from. Our clients are mostly related to AI or data engineering or data analytics or dashboard building. So ml, work is, this is
408 00:50:56.060 ⇒ 00:51:00.861 Awaish Kumar: kind of our in the plan. But right now, we
409 00:51:02.317 ⇒ 00:51:08.989 Awaish Kumar: I don’t think we have any active projects right now, but we can look, you know, like internally, if we can.
410 00:51:09.592 ⇒ 00:51:12.410 Awaish Kumar: Figure out something to work on.
411 00:51:12.610 ⇒ 00:51:14.000 Awaish Kumar: Okay.
412 00:51:14.490 ⇒ 00:51:17.820 Awaish Kumar: But yeah, that’s like, these are kind of like, I, I want
413 00:51:19.400 ⇒ 00:51:22.370 Awaish Kumar: expectations. So like, once you join in and
414 00:51:22.520 ⇒ 00:51:41.750 Awaish Kumar: like we do, maybe we don’t have, like the Ml. Project to work on, or directly. Maybe, like I’m I’m also pushing for some projects which are related to Ml. But but I don’t know how far we are going to get with those, because we already have some things in pipeline which we want to get over before we start something
415 00:51:42.232 ⇒ 00:51:47.040 Awaish Kumar: so other projects which I mentioned, these are the areas where we are working. If you’re interested
416 00:51:48.116 ⇒ 00:51:51.200 Awaish Kumar: like, whether you want to slow on that or not.
417 00:51:52.740 ⇒ 00:51:55.903 Vishnu: So from what you have stated I feel like
418 00:51:57.020 ⇒ 00:52:25.130 Vishnu: you, you basically stated the whole pipeline. And and you said that. Yes, you have a client, a lot of clients who you build dashboards for. And today, today’s day and age, everybody wants to use Llms for whatever use case they have, because they don’t exactly understand it. And they just think it’s like this magical wand that solves all their problems, and they don’t understand that you require compute resources for it. You require a lot of good quality data
419 00:52:25.270 ⇒ 00:52:31.251 Vishnu: for training your model for fine, tuning it and getting outputs out of it. So.
420 00:52:32.140 ⇒ 00:52:51.523 Vishnu: yes, I think the fact that people I mean, if you have clients who are asking you for these kind of solutions. Yes, I’m like super excited to join and like learn again before even joining. I want, I want to ask you this question like, how? How does brain forge operate like
421 00:52:52.620 ⇒ 00:53:11.710 Vishnu: do you like as team members do? Does each one like, can I get on a call with somebody and ask them their thought process behind a certain approach like, is that a thing like, do you do brainstorming sessions where you’re talking about what is a possible solution for this problem statement?
422 00:53:12.870 ⇒ 00:53:19.769 Awaish Kumar: So basically, we have, we actually run our weekly office hours.
423 00:53:20.546 ⇒ 00:53:27.220 Awaish Kumar: Like, where we just come mainly the the that the seniors.
424 00:53:27.550 ⇒ 00:53:37.760 Awaish Kumar: one of the senior engineer is going to be there as a host, and anyone can join, and we have a discussion and brainstorm about thing like whatever
425 00:53:38.140 ⇒ 00:53:44.039 Awaish Kumar: they want to discuss on, like, if they are stuck on something they want. Some ideas want to
426 00:53:44.150 ⇒ 00:54:05.429 Awaish Kumar: have some solution. Just just talk things out like, we have a place for that. But yeah, this is like, kind of a, we want to have a formal place or some something where we want everybody to collaborate with each other. But apart from that, like, obviously in brain forge you are like.
427 00:54:06.576 ⇒ 00:54:14.749 Awaish Kumar: you are like allowed to talk to people. But that’s like, completely depends on on yourself, like like.
428 00:54:15.050 ⇒ 00:54:24.690 Awaish Kumar: how how you go along with your colleagues. But basically we everybody’s like the people who are working. They are collaborative. You can ping anyone.
429 00:54:24.850 ⇒ 00:54:31.850 Awaish Kumar: ask them to come on a call with you on a maybe select hangout or something. And
430 00:54:32.060 ⇒ 00:54:36.009 Awaish Kumar: and then you can basically discuss with them like.
431 00:54:36.730 ⇒ 00:54:43.800 Awaish Kumar: With with leadership leadership team with me, with marketing team sales like you have some ideas, and you want to
432 00:54:44.738 ⇒ 00:54:54.549 Awaish Kumar: discuss them with the marketing team or something like that. You you. Obviously this is possible. We, we actually, we collaborate with each other. We talk to each other.
433 00:54:54.550 ⇒ 00:54:54.990 Vishnu: Okay.
434 00:54:54.990 ⇒ 00:54:57.920 Awaish Kumar: Things like that. But the only thing is that you.
435 00:54:58.640 ⇒ 00:55:18.740 Awaish Kumar: it’s completely depends, like on the on the like. You have a people skills. You, you call on the people. Okay, I want, like, nobody is going to know that we still need something like, it’s some you are going to ping somebody. Okay, I want 30 min of your time today or tomorrow, and I want to have some brand installment session with you. So that’s how it works.
436 00:55:18.740 ⇒ 00:55:22.754 Vishnu: Okay, okay, got it? Got it? Okay?
437 00:55:23.700 ⇒ 00:55:46.750 Vishnu: yes. So I’ll tell you my thought process behind this, because the the reason I asked. This is because, I have been giving interviews for other companies, too, and I have like got like, at least, I’ve got one full time role. And I I’m like thinking whether I should join that or I should join a startup and like all my past experiences, because
438 00:55:47.937 ⇒ 00:56:06.359 Vishnu: I am a fresher now, and I don’t have work experience, I I’m thinking, if I join a startup I will have the flexibility of like, you know, understanding people like I will have flexibility of working across teams. Like again, I I get to work with marketing. I if if at all, that somebody in design team needs help.
439 00:56:06.360 ⇒ 00:56:17.849 Vishnu: I can give my idea. And let’s say, I, I want to track behavior of users, build segmentation, or, you know, run certain experiments about our clients users.
440 00:56:18.578 ⇒ 00:56:33.830 Vishnu: If if you join like a company that has already established, there are already your boss who will tell you what to do, and you you will just blindly sit and Comp. Do those like hit those checklist, and you just give out whatever solution.
441 00:56:33.830 ⇒ 00:56:38.010 Awaish Kumar: Yeah, like at Brent, at Brent Forge. All the ideas are welcome.
442 00:56:38.330 ⇒ 00:56:44.140 Awaish Kumar: We we can like get the we click the idea and then groom it
443 00:56:44.780 ⇒ 00:56:59.779 Awaish Kumar: to do it at a certain level. And if we find out okay, it’s worth spending time on it, and it’s good for for you to learn from a skill, and it’s good for Brainforge as well. That it Brainforge gets something at the end.
444 00:57:00.640 ⇒ 00:57:00.990 Vishnu: Okay.
445 00:57:01.300 ⇒ 00:57:06.149 Awaish Kumar: Yeah, then, okay, we we can work on those projects.
446 00:57:07.470 ⇒ 00:57:08.299 Awaish Kumar: Got it? Got it.
447 00:57:08.300 ⇒ 00:57:18.139 Awaish Kumar: It’s all about like the figuring out. If if it’s worth to spend like 30 days on something if it’s not worth it.
448 00:57:19.860 ⇒ 00:57:21.430 Vishnu: Got it. Okay, okay.
449 00:57:21.730 ⇒ 00:57:28.341 Awaish Kumar: So yeah, so that’s the plan. And yeah, we would like, I would just would update you on
450 00:57:29.030 ⇒ 00:57:37.349 Awaish Kumar: on when we can like, if, like, like, I go back and discuss with the team, and we can decide on
451 00:57:37.490 ⇒ 00:57:45.500 Awaish Kumar: whatever our decision is. And if if it’s and whatever it is. And then what is the plan? If it’s a yes, then
452 00:57:45.900 ⇒ 00:57:50.439 Awaish Kumar: how we are going to move forward with internship, and also like
453 00:57:51.248 ⇒ 00:57:54.250 Awaish Kumar: And then obviously, you can also do.
454 00:57:54.380 ⇒ 00:57:59.250 Awaish Kumar: Think in the meantime, and give us the feedback.
455 00:57:59.460 ⇒ 00:58:02.117 Vishnu: Yes, one more question like
456 00:58:03.439 ⇒ 00:58:13.970 Vishnu: let’s say, after the internship. What are the odds of this internship converting to a full time like is there even a full time role at Brainford?
457 00:58:14.430 ⇒ 00:58:16.330 Vishnu: like, how does it work.
458 00:58:16.710 ⇒ 00:58:23.270 Awaish Kumar: So at Benford, we basically, as I mentioned, with consultancy. So there, there are good chances that
459 00:58:23.610 ⇒ 00:58:46.820 Awaish Kumar: so if if we have someone and as an intern with us, he works on analytics, engineering. We have like, spend our time to mentor someone to learn and and create some. A. A. DVD. Projects and then work on, for example, some warehouses, and then build dashboards and also build some pipelines. Then we would like to keep
460 00:58:47.080 ⇒ 00:58:53.480 Awaish Kumar: keep him in with us. If we have some projects, it totally depends on how many clients and how many
461 00:58:53.630 ⇒ 00:58:57.749 Awaish Kumar: oh, projects we have so.
462 00:58:57.750 ⇒ 00:58:58.350 Vishnu: Okay.
463 00:58:59.720 ⇒ 00:59:05.411 Awaish Kumar: But like it, it like it. Obviously it is totally dependent on
464 00:59:07.110 ⇒ 00:59:13.289 Awaish Kumar: on on the need like, and also the what you say, like.
465 00:59:13.540 ⇒ 00:59:25.249 Awaish Kumar: like, if we see the potential in a in in a few weeks. And we so we have an opportunity to like someone to hire like. Then obviously.
466 00:59:25.600 ⇒ 00:59:31.019 Awaish Kumar: that person is going to be the 1st choice to like. Be a full time.
467 00:59:31.270 ⇒ 00:59:32.380 Awaish Kumar: Come on with us.
468 00:59:32.380 ⇒ 00:59:33.924 Vishnu: Got it. And
469 00:59:34.900 ⇒ 00:59:45.864 Vishnu: again, this is this is, you can decide not to answer this question also. But I want to understand like has brain forge raised funds recently, like
470 00:59:46.970 ⇒ 00:59:54.659 Vishnu: from any investor from a Vc or what is like. I’m not sure it’s like.
471 00:59:55.210 ⇒ 00:59:58.605 Awaish Kumar: Yeah, it’s a like good step as
472 00:59:59.290 ⇒ 01:00:01.290 Awaish Kumar: startup, and we are mostly running.
473 01:00:01.630 ⇒ 01:00:05.929 Awaish Kumar: Running with with our own revenue, which we get from our clients.
474 01:00:06.580 ⇒ 01:00:07.569 Vishnu: Okay. Okay.
475 01:00:08.250 ⇒ 01:00:34.378 Vishnu: okay. Okay. Well, 1 1 last question, like, let’s say you’re in my shoes. And you’re like, clearly, like, I have gone through your Linkedin, you you have so many years of experience, and so I’m sure you know how the industry works. And now I am somebody who who is just fresh out of college. And I have just like, if you look at my altogether experience. 3 years, 2, 2 and a half 3 years of experience. So
476 01:00:34.920 ⇒ 01:00:44.016 Vishnu: if you are in my shoes and like you have this opportunity where you’re joining a startup for an internship, and you have this full time offer from
477 01:00:44.780 ⇒ 01:00:48.089 Vishnu: from one of one of the S. And P. 500.
478 01:00:48.090 ⇒ 01:00:58.170 Awaish Kumar: I’m not not in a position to answer that. Like. It’s it’s completely your choice like, I have made my own decisions in my
479 01:00:58.310 ⇒ 01:01:08.970 Awaish Kumar: Ted, you know, like to work with Startup at some points like, but I cannot answer it for you like what is the best for you? I don’t know. I don’t know your circumstances. I I just let you in this meeting, and I don’t know like.
480 01:01:08.970 ⇒ 01:01:09.320 Vishnu: Yes.
481 01:01:09.320 ⇒ 01:01:13.772 Awaish Kumar: What your situation is, what you’re looking for. So it’s completely on you.
482 01:01:14.360 ⇒ 01:01:16.969 Awaish Kumar: maybe we have you. You have the
483 01:01:17.491 ⇒ 01:01:21.108 Awaish Kumar: time to explore. Look what Redforge is doing, and
484 01:01:21.830 ⇒ 01:01:27.690 Awaish Kumar: you, you are better. But, like you are the person and like you know
485 01:01:27.900 ⇒ 01:01:31.260 Awaish Kumar: your life better than me, and you are the.
486 01:01:31.260 ⇒ 01:01:31.660 Vishnu: Okay.
487 01:01:31.660 ⇒ 01:01:33.590 Awaish Kumar: A better person than me to answer that.
488 01:01:34.550 ⇒ 01:01:35.750 Vishnu: Got it. Okay?
489 01:01:36.120 ⇒ 01:01:41.649 Vishnu: Okay. Oh, yeah. Okay, thank you so much. For, like, taking time to do this.
490 01:01:41.650 ⇒ 01:01:46.470 Awaish Kumar: Thank you. We are. We are just over the time, and it was nice talking to you, and we.
491 01:01:46.470 ⇒ 01:01:47.250 Vishnu: Same here, same.
492 01:01:47.250 ⇒ 01:01:48.530 Awaish Kumar: So thank you.
493 01:01:49.120 ⇒ 01:01:50.779 Vishnu: Yeah. Have a good one. Bye, bye.