Return to site

80. Unravelling the Data Galaxywith Nitin Singh - Part 1

· AI,podcast


broken image


Podcast with Nitin Singh Part 1


Summary: In this episode, join AndrewLiew as he chats with Nitin Singh, a data science leader with over 15 years of
experience. Nitin shares his inspiring journey from IT to data science, his
passion for continuous learning, and his invaluable insights into designing
effective data science solutions.

Nitin Singh, an experienced datascience leader, recounts his unconventional path from IT to data science.
Starting in India's competitive education system, Nitin found his passion for
data science and embarked on a journey of continuous learning. He shares the
importance of having a supportive friend circle and constantly pushing oneself
to excel in one's field. Nitin emphasizes the significance of solution design
skills in the data science domain and the need for top-to-bottom learning and
hands-on projects to truly grasp and apply concepts. His passion for continuous
learning and attention to detail have made him a distinguished data science

[00:00:00] Andrew Liew: Hi everyone. Welcome to the show.Thanks for listening to this episode with Nitin Singh. Today, my guest here,
Nitin has over 15 years of experience in leading teams and delivering data
projects. It's currently serving as the head of data science and engineering at
Molladin, a rapidly expanding fintech startups.

[00:00:21] Andrew Liew: His deliverables includedeveloping language and visual models for car inspection automation, creating
data driven pricing frameworks, and implementing text to text generative
models, with his previous role as the Director of Data Science at Gartner, an
AVP of data sciences at WNS. Nitin has let teams set up data science practices,
develop, deploy models for numerous critical business use cases, and receive
prestigious industry awards such as the gold medal at the Stevie Awards and the
Nescom, the Top 50 AI Game Changer Award.

[00:00:58] Andrew Liew: Apart from that, he has [00:01:00] won Leadership Awards and GlobalInnovation Continuous Improvement Award at Gartner. Nateen holds a postgraduate
diploma, MBA from the Institute of Management of Calcutta, and a Bachelor of
Engineering from Natarji. Subash institute of technology along with certificate
in deep learning ml pipelines on google cloud docker's containers and aws
practical, data science among others.

[00:01:25] Andrew Liew: So welcome to

[00:01:26] Nitin Singh: the show Thanks a lot and you'rereally honored to be here.

[00:01:28] Andrew Liew: Yes. I actually a few of myguests actually reach out to me to say hey, you gotta talk to nitin. I a very
cool guy Ping up and you came on to the show. So tell the audience like about
your back story. Apart from me reading out your profile, they want to hear from
you, from the actual person.

[00:01:46] Andrew Liew: How do you became a data scienceleader at Molodin from the day you finished school at IIM Calcutta?

[00:01:50] Nitin Singh: So, it was never a plan, to behonest, and being from India, the kind of education system we have, it's more
of a generic you tend to study a lot of things [00:02:00]and till my graduation or even till my undergraduation, I was never sure what
exactly I'm going to do.

[00:02:06] Nitin Singh: So the basic, the first, thesafest bet is in Indian after colleges is to get a job in I. T. Pretty
straightforward. And you're sorted. But some so, that exactly what I did as
well. But a lot of my friends during graduation, they did MBA as well. They got
to, they went to do, GRE, then MBA from GMAT from U.

[00:02:26] Nitin Singh: S. They went to different placesacross the world. So the thing I want to tell you, everybody is that it's very
important to have a good friend circle. Yes. The friend circle actually defines
you because your average of six people you meet. And that was the point for me,
because after college, I did not understand what I was doing.

[00:02:43] Nitin Singh: I was not. I did not know whatI'm doing, what I should be doing, but somehow I realized that what I'm doing
right now not something I really want to do. And then I spent three or four
years. Doing really well in IT industry. I was a DBA and then I was a full
stack developer and end up being a full stack [00:03:00]developer because I was never fine with just knowing one thing.

[00:03:03] Nitin Singh: I always want to exploredifferent things. But anyway, I have a younger brother, so I used to teach him
for, CAT, that's MBA competition, sorry, MBA entrance exam in India. It's a,
it's like the best colleges in India, the IIMs.

[00:03:16] Andrew Liew: Yeah, before you talk about,helping your brother let's talk about, you mentioned about the friends, and you
mentioned that quote you are the sons of your six friends.

[00:03:24] Andrew Liew: It just draws me the code when Iactually met Drew Houston, the founder of Dropbox before he said the same
thing. And so coming from that, they say that at a time when you graduate, you
don't know what you want to do, but you have friends and eventually everybody
seems to be heading to this IT or info technology or info communication

[00:03:43] Andrew Liew: But IT is so broad Who are your,some of your good friends that you spent a few hours chatting about live and
then you end up in your first job.

[00:03:51] Nitin Singh: So one of my friend is heactually is he's doing a startup in us in California. I'm sorry, San Diego. And
he was a bright chap, a very good [00:04:00]friend of mine.

[00:04:00] Nitin Singh: And he was, so he went to us andhe, told me that he's only going for a couple of years. He'll just do his MS
and come back. He is never back. He's been 15 years now. , and another friend,
I think a very good friend of mine. He basically he was also from a semi
country industry. He was in Texas Instruments.

[00:04:16] Nitin Singh: So I was, and another friend thethird friend of mine other. He actually right now is in U. S. He's in
California. He actually after college did M. B. A. From I. M. He got an I. M.
Lucknow. So I was feeling what exactly I'm doing when I talk to them, at least
even if they don't know what exactly they want to do, but they are in a field
where They're good.

[00:04:37] Nitin Singh: Okay. See, the only thing whichdefines me is I want to be at a place where I'm really good. What I'm doing is
something not anybody can do. So if I'm not differentiated, if I'm not my work
is not something which is unique for my company, then I can be replaced easily.
So that's what keeps motivating me.

[00:04:53] Nitin Singh: So when I talk to my friends. Irealize I need to be good in something, which is I need to be really good in
something. [00:05:00] That's how the wholejourney towards, even till 2011, I had no idea what exactly I would be doing
that, especially during I'm calta Yes. When I was studying statistics and when
in 2011 the deep learning actually broke.

[00:05:12] Nitin Singh: We have a, major competitionbreaking the benchmarks. That's what motivated me because statistic was
something I really like always. Oh, so nice.

[00:05:21] Andrew Liew: Is it because you, I don't knowwhether you play poker or some gambling games and then your friends say, Hey
Natine, you're so good at winning me because you're good at statistics and
probability is that case.

[00:05:31] Nitin Singh: No Not, like that. See so, theIndian system is very initially when you study you see lot, of things. You
study a lot of things. Yes. And statistics was one of them in my undergrad. But
during M B A when lot, see, I'll tell you what, happens is Iams, okay? Yes, we
have lot of good tech, technology colleges.

[00:05:49] Nitin Singh: Its. Yeah, and it's very,difficult to get into them. I

[00:05:53] Andrew Liew: mean, it's, I heard it is socompetitive. India got such a big population, right? We're talking about more
than a billion, [00:06:00] right? And then evenevery college would probably be like a few in the millions, right? Your
students would have probably feel the entire population of singapore.

[00:06:08] Andrew Liew: We only have six

[00:06:09] Nitin Singh: million people I tell you what Itell you what people actually Prepare for whole year two years to get into
those colleges Wow You know what? One more important thing I want to bring to
the light is. You will see not many top colleges are in top hundred, right?
When you look at the global.

[00:06:24] Nitin Singh: Yes. The reason for that is whenyou look at any ranking, it actually is very it's, a spectrum. You look at
multiple things. Now, infrastructure, I agree may not be as great as some
colleges in Europe, probably, yes. But the only one thing which is focus which
is which people actually really focus on is the aptitude.

[00:06:43] Nitin Singh: So imagine somebody studyingphysics, chemistry mathematics for two years and, then giving a competition
it's very hard to clear. So in IIMs, what I'm trying to tell you in IIMs, a lot
of people from IIT comes in. And IMS actually came after I ts and the lot of
culture of IMM is [00:07:00] actually derivedfrom IOTs only.

[00:07:01] Nitin Singh: Ah, interesting. Yeah so, when Ientered when I got into i m, okay, first of all, I was very lucky to get into
it. I was basically teaching myself, teaching my brother and I cleared, he
cleared some of the college. He did not clear I m, so I was very lucky to clear
but when I got into, believe me, it Delhi, it, Mumbai a lot of people, talented
folks 10 pointers.

[00:07:22] Nitin Singh: When I got into IAMs, I realized450 people and each and every one of them is really good. Somebody from from
Singapore, a lot of people come through GMAT, they got into IAM Calcutta,
people who are, vice president Goldman Sachs, they are there 10 pointers from
IIT Delhi.

[00:07:39] Nitin Singh: So I was really I realized, dudeyou're not the compared to all these people, you're not something. You cannot
compare yourself, man. They're like damn talented people. So that's something
which I realized very early in my life is that there's always somebody who's
better than you, okay?

[00:07:55] Nitin Singh: Don't think yourself as somebodywho got into a college. On cloud nine, you're the [00:08:00]best.

[00:08:00] Andrew Liew: Like you, you probably realize there'sso many like superheroes, right? Like Avengers. So then how do you discover
your own superpower? What is your unit? Because you see so many heroes, so many
intelligence people, you

[00:08:11] Nitin Singh: know, see one thing, one thing Irealized, no talent is not something which is just scoring marks in exams.

[00:08:18] Nitin Singh: Yeah. Okay. Talent is a spectrumand somebody who is actually good across the spectrum. Even though they may not
be best in one particular area, if they have a good spread those are the people
who do really well. And that's what I've noticed. Somebody who scored 10
pointer, but who doesn't know how to communicate himself.

[00:08:35] Nitin Singh: Those people generally faceissues because see, when you join a company the company doesn't look for a
superhero. They look for, someone who can adjust to their culture.

[00:08:44] Andrew Liew: Like a superman, right? He's justa journalist,

[00:08:47] Nitin Singh: right? Exactly. But some peopleare there and some people are, who would just study for a couple of hours
before the exam.

[00:08:54] Nitin Singh: And then they just score marks.Like anything they won't attend the classes. I've seen those people and I
seriously [00:09:00] can't imagine how they'reable to do that. The same thing, I studied for eight hours, so I have my
friend, okay. He actually in Singapore only, and he's currently head of
finance, corporate finance head of investment.

[00:09:10] Nitin Singh: Okay. Vietnam based so in, in,Iams. Okay. Yeah. And he was also placement representative. Okay. Basically, he
used to communicate with companies and See, before the exam, he always would be
busy with talking to companies and all his attendance would also be the bare
minimum. And we would study for exam and he would come one day back.

[00:09:29] Nitin Singh: He was, he will ask us to justteach me for a couple of hours and this for more months. So those are the
people who are really talented, what I've noticed. But yeah you, know what, the
way I realize, because I need to focus on my strength and that's very
important. And it's not a strength, it's what keep you passionate, what you're
passionate about, especially in data sciences.

[00:09:48] Nitin Singh: It's very difficult for anybodyjust to do data science because it's in trend. Data science is not just data
science, it's a big field, to, for, I'll give an example. A basic, model, which
we call [00:10:00] linear regression. Yes. Toknow linear regression, you know the cost function. You know our

[00:10:05] Nitin Singh: Maxima, minima. Matrix. Linearalgebra. Statistics. Hypothesis testing. P value. What is p value, z value. You
know what is chi square test. If you actually... Breakdown the linear
regression. There are many concepts.

[00:10:16] Andrew Liew: Yes. I don't know.

[00:10:18] Andrew Liew: There was always some junior datascientists just say it's just a loss gradient function or XGBoost. That's just
a package, right? You got to understand the need it's like you drive a car, you
need to understand the engines, the dashboards. So you, when you drive, you
know where to move the gear, right?

[00:10:32] Andrew Liew: In the same way. So coming backto you so you move into like full stack developer, and then later, what
happened next, you

[00:10:39] Nitin Singh: know? So got into MBA. Now afterMBA, You and people would normally manage people, right? Yeah. Master of
Business Administration, you manage people. So I joined, I had a couple of

[00:10:50] Nitin Singh: One was from a sales in US,California. California is somehow very close to me, man. A lot of things are
happening. So I got an offer, one from California. [00:11:00]Okay. The other one was from a IT based firm. A very good firm from Delhi. Oh,
Delhi, okay. At that time I was 29 and I was not married. Ah, nice.

[00:11:10] Andrew Liew: Okay. Your full

[00:11:12] Nitin Singh: Yeah, Amazon. I at the same timeI wanna get mad obviously, and also I did not have any, I never really wanted
to go outside to be honest so I, was not very ambitious to start with at that
time, okay. I just wanna do very well. That's about it. I never thought that
I'd go outside India and do something.

[00:11:28] Nitin Singh: So I got married and then. Istayed for like around five years in a there's a firm called Nagaro. Okay.
Okay. It's a Japanese name. And this firm has a good business in Europe, US

[00:11:40] Andrew Liew: Oh, were you taking on that?Yeah.

[00:11:43] Nitin Singh: So I joined, that for a man thebest part, see these firms are very underrated because what you learn there,
you cannot learn anyone, Microsoft or Google.

[00:11:51] Nitin Singh: Yeah. See, turn back time.Actually, when I came to Singapore, I actually declined a Google offer. I,
anyways back to you

[00:11:58] Andrew Liew: mentioned you were at this [00:12:00] Japanese firm and then like what, whateventually lead you to another chapter in another part? Continue please.

[00:12:07] Nitin Singh: Yeah. So. it's not a Japanesefirm, it's an Indian firm, but the name Nagaro that's, Oh,

[00:12:11] Andrew Liew: okay.

[00:12:12] Andrew Liew: Okay. So the Indian firm, but itsounds like a Japanese. Yeah.

[00:12:15] Nitin Singh: So I learned a lot from therewhen you talk about how to manage teams. So we had the great managers. We have
great delivery managers Who actually I look up to them and I, that's how I, got
to know how to manage teams, how to create a product and how to lead teams, in
fact, how to manage risk in a project, quality in a project, right?

[00:12:35] Nitin Singh: Yeah. At the same time there wasthat thing going in the back of my mind, the machine learning regression and
all those things. Yeah. So that was a time. I used to study every night from 12
a. m. to 3 a. m. in the night once my wife sleeps.

[00:12:50] Nitin Singh: So there's a guy calledShibirani, just Stanford. There's a Stanford course. By these two legends man.
So I, basically went through that course around [00:13:00]15 to 20 times and everything went, through those their, course. It was
amazing. I always learn new things. And that is something I believe right now
it's very common to find good good resources for learning technology machine

[00:13:14] Andrew Liew: know.

[00:13:14] Nitin Singh: 2014, I'm talking about, yeah,those two guys, a Stanford professor, their course is brilliant. I still would
do wanna do that again, and that's so in the duration of four years or five
years at Nagarro every night, I used to study for 12 to three hours and I was
very passionate about it because I realized this is something where I can be

[00:13:33] Nitin Singh: Yeah, this is something where Ican add value. And my nagar was also very generous to let me explore in that
area as well. So after a couple of years, I then started focusing on projects
which were based on machine learning. A lot of p o c we did at that time,
Hadoop was also there.

[00:13:52] Nitin Singh: Is

[00:13:52] Andrew Liew: Nagarro a consulting firm or Itwas in Oh, it was. Okay. Okay. Continue.

[00:13:57] Nitin Singh: Yeah. We do a lot of projects aswell. Build [00:14:00] application andproducts, but mainly it was yeah, it's a consulting firm only. So that's how it
all started. And, After Nagaro, I wanted to do something which is very,
individual contributor based because I really want to keep my hand dirty and

[00:14:13] Andrew Liew: say because most people like whenthey reach their 30s when they are a manager or leader they want to move and
improve their people skills where you took a unique path or a narrow path where
You decided to be more technical, right?

[00:14:25] Andrew Liew: Tell me more about what were youthinking? How do you decide to take that alternative path, which is very
different from those who are following at that time,

[00:14:34] Nitin Singh: see, it is very important for youto plan your career. Yes. Okay. You cannot be in a field where what you are
doing, you can be easily replaced.

[00:14:42] Andrew Liew: Yeah, these days got so many, I

[00:14:44] Nitin Singh: got a lot of things would bereplaced, but I think the main, the major thing that would be replaced would be
the work, which is painful for us to do. Those things will be automated.
Automated. But yeah. So I always focus on your career, like next five years,
next two years, what you want to do.[00:15:00]

[00:15:00] Nitin Singh: Yes. And at my first companyNagaru, I realized. What I'm doing, it can easily be replaced at anybody can
manage project. Anybody can manage, but that was not the case. We'll be honest.

[00:15:11] Andrew Liew: It's I work in big companies,small companies, government agency, actually people's skills can be underrated.

[00:15:18] Andrew Liew: But even though you have verygood communication and people's skills. What was that driving impetus saying,
Hey, I need to, I want to be more technical. I want to be a more like a top
leader, individual contributor. Like how do you eventually move to that path?
And it also requires sacrifice, right?

[00:15:34] Andrew Liew: I'll

[00:15:35] Nitin Singh: tell you what. So when I talk toclients when I talk to somebody, let's say a technical architect from a New
York based firm, where we are doing business. When I'm talking to that person,
I don't want to, go back and talk to my technical architect, Hey man, he was
proposing this architecture is, would that work?

[00:15:53] Nitin Singh: I don't want to do that. I wantto, when I'm, so it's, my personal thing. Okay. So I, there's something called
a constructive ego[00:16:00] to that. So Yes, Ihave a constructive ego because I want to know what other person knows. Okay. I
don't want to be I don't want to feel myself that there's something which I
don't know in my field and somebody else know.

[00:16:11] Nitin Singh: That's how I keep motivatingmyself. And I believe, and there's a fear also, there's a constructive fear as
well. It's my, construct. So I feel if there is something If there is something
to look up that I don't know, then I'll drive myself to learn it. And when I'm
learning it, then I'll be like more comfortable.

[00:16:28] Nitin Singh: Okay. Okay. For example, the whenthe bird and you will transform a game in. Now, I know the concept using birds
all the previous concept in. LSTM pretension positional encoding, then you
know, then you have a encoder decoder architecture.

[00:16:42] Andrew Liew: We share that when you talk aboutBERT or when you're associated using BERT.

[00:16:46] Andrew Liew: Yeah.

[00:16:47] Nitin Singh: Yeah. So when it came, so if Idon't know how internally BERT works, if I internally, so BERT is only encoder
part of a transformer, it's not a complete encoder decoder, it's only encoder
part. Yeah. If I don't [00:17:00] know how youtrain Bert, how the Bert is trained, it's in multiple ways, like classification
and sentence comparison you need to know all these things.

[00:17:07] Nitin Singh: You need to know how you put allthe data into the Bert and how the, how you can use the output. What is the
difference between the embeddings you get in Bert compared to lecture work to
work, right? So if you don't know all these things, if you don't know all these
technical details, you cannot actually come up with a good solution.

[00:17:23] Nitin Singh: Now, in data science, the mostimportant thing is the solution design. Especially right now, right? Like in
technology, you have a technical architect, so in data science, you need to
have a good solution design skills because I think model creation has been
automated to a large extent. You can convert one shot learning.

[00:17:39] Nitin Singh: A lot of things are there. Youdon't need to create your own model. The challenge is actually to design. See,
model is only probabilistic, right? Model and the business required
deterministic solution. Yes. So if your model is 90% accurate, you need to
design it in such a way that the 10% error is managed in a good way.

[00:17:56] Nitin Singh: The consumer doesn't have to dealwith that error. That's [00:18:00] where thesolution design skills comes in. So that's very important. And that is the
reason you know, I really appreciate it. Want to focus on technology and
believe me after this, I also learned the data engineering, the backend. I know
when I have to share the models at that time, like 10 or five years back, share
the model with my leaders.

[00:18:16] Nitin Singh: I learned Django. I created manyapplications shared with the leaders right now. It's very easy. You can have
radio, a lot of you can have a flask a lot of things are there, but you need to
keep learning. If you don't. Keep learning that it's very difficult to be a good
solution design or a data scientist for that matter in machine learning.

[00:18:35] Nitin Singh: So that's very important. Oh,yes. I

[00:18:37] Andrew Liew: totally align with you that inthe world today is not about know about learn or learning so many stuff,
learning so many branches. Interesting question that probably the audience will
want to know is like, how do you. pick the direction or where you want to learn
even though your sequence you move from full stack to machine learning to NLP,
natural language programming, and then move down further to the mechanics [00:19:00] of the encoder and move down to the data.

[00:19:02] Andrew Liew: So that's, a sequence, but thereare some people like when, they are fresh graduate they were like, wow, there's
so many things to study. What is the best, Approach or what is your natural
inclination to find out where's that good learning

[00:19:15] Nitin Singh: direction? It was very tough tobe honest forget about deep learning.

[00:19:18] Nitin Singh: So in 2013, 2014, I was not apython developer, but every time I used to think of python. So I used to find
multiple packages doing the same thing. I always would be confused. How do I
learn this language? The many ways to do the same thing. And then I realized
there's a concept called top to bottom learning.

[00:19:36] Nitin Singh: I always focus on bottom toupward learning where you focus on details. And then you finally. You become
know everything then you cannot know everything anytime. Okay, never You can
only try to strive for Being the best but you should keep trying that you know
that point you cannot you can never read So what I realized that when you're
studying something, I think continuity is the main thing [00:20:00] even if one day you're spending three hoursJust try next day.

[00:20:03] Nitin Singh: If you're busy, try to at leastspend half an hour or 15 minutes. That continuity is very important. Once you
lose that continuity, all the effort goes wasted and it's always best to write.
Okay. I always spend a lot of time in my school where I never used to write and
you just used to look and try to things I know everything.

[00:20:21] Nitin Singh: But if you're writing the mostsenses involved, so it actually helps you remember better. If you write what you're
learning, keep making the notes. I used to go back to the main book many times.
Then I realized why not just create my own notes only. So the notes which I
created in 2014, I still use it a lot of times.

[00:20:38] Nitin Singh: Algebra, mathematics. I still goback and look at those notes. So it's very important because mathematics is not
changing. Yes. Would not change to five. Okay. So continuity is very important
and if you think and bottom to top to bottom you learn anything while doing a

[00:20:54] Nitin Singh: Okay, don't go by book chapterone chapter two you'll, suck in that book for a long time. [00:21:00] And I'm is running very fast, start to doa project. So in a project, let's say it's a regression problem, you're solving
a regression problem, you understand how to make prepare data, then you
realize, okay, how do I find attribution.

[00:21:11] Nitin Singh: Then you see, okay, how do I knowit's the best model? You look at the output and say, okay, this p value is not
significant. Hey, what do you mean by p value? Then you go and search for it,
study it. Then as soon as you search for p value, you have to learn hypothesis
testing. When hypothesis testing, you say, okay, it's a t test or it's a z

[00:21:27] Nitin Singh: Okay, that's a sample. Okay, it'sit's continuous data. Hey, what about it's a categorical data? Okay, there's a
high score test as well so that's how you start learning a lot of things which
are more relevant to your work, right? So

[00:21:40] Andrew Liew: coming back to backtrack a bit.So once you eventually move into ABP data science and WNS, what is it that you
realize now you have the technical skills or you realize that you only just
know the beginning?

[00:21:51] Andrew Liew: How does it work? So you movefrom. People management skills, you become an individual contributor and how
the story unfolds

[00:21:57] Nitin Singh: yeah, so after [00:22:00] Nagaro, I went to the publicity group, abig company. There I was, I at least rolled out around 40 plus market mix
models, structure, equationing models, equation model, attribution analysis,
cost optimization.

[00:22:12] Nitin Singh: So that was a great experience.The team actually was dismantled after. Six months. I think my boss also left
that firm, but in those six months, I learned a lot. Brilliant. I actually,
okay. There's something very interesting. Yes. Yes. Especially with linear
regression. Okay. So see a linear regression model have multiple solutions.

[00:22:29] Nitin Singh: Yes. Right now, how you actuallyhow it happens in business. So you talk to the business. Okay. Hey, this is a
product. What do you think is the best channel where you get the most return
that would say, okay, see, I'm very sure about this Facebook ads. And I'm very
sure about is out of home promotions.

[00:22:47] Nitin Singh: Now, when you go back, you do themodeling, you ensure the P value of those two channels are high. They have a
good weight in the output. So you can go back and say, okay, yeah, you are
right. You're good. Feeling was right. So two, [00:23:00]two things we are fine. They are important. Other thing we'll figure it out.

[00:23:03] Nitin Singh: It's different weights and all soit's actually. The way analytics has been, it's more driven by the business
side because business wants to validate what they're thinking is right or