Return to site

81. Data Science and Leadership in Startup and Corporate Environments with Nitin Part 2

· podcast,AI


broken image


Podcast with Nitin Singh Part 2



In this podcast, Nitin Singh, a datascience expert with experience in both startup and corporate settings, shares
his insights into the challenges and benefits of data science in different
environments. He discusses how leadership plays a crucial role in driving data
initiatives and fostering a culture of innovation. Nitin also compares the
nimble and fast-paced nature of startups with the structured processes in large
companies and emphasizes the importance of continuous learning and adapting to

In this engaging podcast, Nitin Singh,a seasoned data scientist, reflects on his experiences in both startup and
corporate worlds. He highlights how leadership can make or break data science
initiatives and discusses the critical role of data accessibility and
infrastructure in driving innovation. Nitin shares how the startup environment
fosters quick decision-making and nimble strategies, allowing for faster
implementation and adaptation. In contrast, large companies require structured
processes and change management to execute data initiatives successfully.
Throughout the conversation, Nitin emphasizes the importance of continuous
learning and remaining adaptable to stay ahead in the dynamic field of data

[00:00:00] Andrew Liew: I often encounter, even myself,it's called the curse of the knowledge so much suddenly when you talk to the
business leaders, you need to learn to I've struggled to dumb down into simple
business language because the guy like, so this linear regression model or
marketing mix model, they were like, So what does that mean for me? How do you,
handle that? I don't know. Do you ever encounter that challenge in the early
days? When you. A lot. Yeah.

[00:00:22] Nitin Singh: A lot. I think this is where thechart GPT actually helped. Even, this year before, before chart GPT came,
people were still skeptical about deep learning, especially it's a black box

[00:00:33] Nitin Singh: They don't understand withregression. So regression is still the most widely used algorithm because you
can find the attribution directly. Yes,

[00:00:40] Andrew Liew: it can. Yeah,

[00:00:41] Nitin Singh: very easy to explain to businessyou can tell that this is what's impacting your outcome and how you can
actually improve that outcome, right?

[00:00:49] Nitin Singh: By applying different strategiesand all. And you can again run that. So it is very easy to understand, but deep
learning is very difficult to explain because you cannot directly [00:01:00] attribute anything. It's all weights andall right, it's all neurons and all you cannot do that. But after ChatGPT,
people have somehow start trusting, looking at a positive side of machine
learning models.

[00:01:10] Nitin Singh: Okay. It was free first. If itwas not free, this would not have happened. Initially.

[00:01:14] Andrew Liew: Premium actually works, right?The

[00:01:16] Nitin Singh: business model. It definitelyhelped because now business is more. Understanding towards power of this
machine learning model. Earlier, they would say if you give them accuracy of
99%, they would say, what would happen to that 1% case?

[00:01:29] Nitin Singh: I cannot have my business beingthat one person error. I don't want, and business always would want complete
surety of outcome. They want, so they want deterministic solution to a
probabilistic problem. So that is something which is contradictory because it's

[00:01:43] Andrew Liew: human nature. I like they, wantto share things to put your money, even though they know it's almost like
buying a lottery ticket, maybe a more calculated investment.

[00:01:51] Nitin Singh: It was very difficult. All right.Initially very difficult to explain business. So we would explain, so we would
show them, they can see, this is your [00:02:00]training set, we train the data on these just like a normal kid would
understand, look at things. That's how the model is learning. And this is the
data we hit and this is.

[00:02:08] Nitin Singh: We tested on, and this is theaccuracy of business would not buy that business would say, and they should be
because you're not sure the distribution, which you use is actually the
complete distribution, right? So there's a lot of assumptions in the model we
made, but the best way to actually, I'll tell you the best way how to do it the
best way we implemented.

[00:02:27] Nitin Singh: I learn that from Gartner inprevious companies as well. We used to do but not as effectively as in Gartner
Not a lot of people misunderstood Gartner as you know They provide a lot of
research reports and all in ML.

[00:02:38] Andrew Liew: I'm curious Please tell us likewhat exactly that you do in Gartner apart from people having this Perceived
notion is just a a Moody's review on the tech products

[00:02:48] Nitin Singh: Yeah, see in Gartner.

[00:02:50] Nitin Singh: We do a lot of So right. I'm notpart of that now, but

[00:02:53] Andrew Liew: what do you guys exactly do?

[00:02:55] Nitin Singh: So we basically provideintelligence to our salespeople, [00:03:00]communicate with the business, the CFOs, CIOs. All the intelligence which they
can use and do the retention. So best part about Gartner is everybody in the
whole organization is very clear what their metrics is, what the North star is.

[00:03:13] Nitin Singh: So we know retention is somethingwe have to focus on, right? I never find this clarity in any companies, to be
honest. They are very, clear what they want. Everybody, till the new developer
joining the company or a new business guy joining the company, they would know
what the main metrics we have to focus.

[00:03:28] Nitin Singh: Wow. That's the beauty of thatfrom amazing from now in data sciences they are the first people. The leaders
are great. Leaders are very, much data oriented. Believe me. I had one MVP when
I joined Gartner. She was so good. She was from IIT Madras. She was so good in
mathematics that she would actually understand what model you made and she
would also give you ideas.

[00:03:50] Nitin Singh: So believe with a lot of peopleyou know who are very good in technically as well and they are equally good or
better in managing people. So these two skills can be. Can be with a person at
the same [00:04:00] time. Okay. It's not a zerosum game. That's another thing I've learned, right? So the, leadership was
very, accepting towards these initiatives and they would put all their effort
to ensure that we have the outcome.

[00:04:13] Nitin Singh: Now, I was telling about yourway, how to execute those projects. So what we would do in Gartner, we would
basically create a pilot. In pilot, we will define the problem. We will have a
hypothesis and we'll have a matrix. Okay, we'll use the models. We'll do
backtesting. We'll do live testing with business user as well and look at a

[00:04:32] Nitin Singh: It would be a very small durationproject. Let's say three or four weeks and three or four weeks. We'll see
whether all the effort all the model is actually helping improving that matrix.
If that model is not helping. That matrix to improve, then we'll snap it off

[00:04:47] Andrew Liew: three to four weeks for aproject.

[00:04:48] Andrew Liew: That means your scope must belight or tight, right? It can't be like running on Petra bike data or multiple
thousands of columns of variables, correct? I don't know.

[00:04:57] Nitin Singh: No, Because see, Gartner is very [00:05:00] data rich. So that's what we get. It'svery data rich and they have their own platform where people across industry
come at.

[00:05:08] Nitin Singh: I'll tell you what, the power ofdata which Gartner has. So when did, so I think during COVID. Yes. There's a
thing called work from home. Some term came in actually, people working from

[00:05:19] Andrew Liew: home. Yeah. Work from home, RTO,return to office. Yeah.

[00:05:22] Nitin Singh: So we actually know that term.Even before it came into the market because see Gartner is first one to know
when people across the industry and ask questions to our platform, right?

[00:05:31] Nitin Singh: Yes. So we are very much ahead.The Gartner is very much ahead and understanding the market situation because
they have access to top 500 fortune companies. What's happening in the market
because people come there, ask questions, look for reports, talk to a
consultant, right? Yeah. We have that advantage, which nobody in the industry

[00:05:50] Nitin Singh: So in

[00:05:51] Andrew Liew: other words, Gartner must have. Afew years ago, 10, 20 years ago, have that vision of let's build a data
strategy right to capture all the data in order for you to [00:06:00] quickly turn around. I may have access tothe data and flip a model quickly, right? Because when I was working in a big
company, even big consulting firm like with the McKinsey's or the conference
man, taking the data, depending on this vendor, that vendor, that storage, that
access that's fair.

[00:06:14] Nitin Singh: See, they have data stewards.Okay. You won't find the MDM in other companies. People talk about it. They
talk about master data management and all. Okay. They have a, so Gardner data
is high quality. Teams are very good. The data engineers are also that far. I
was surprised to see so much talent in a firm.

[00:06:32] Nitin Singh: And in data sciences we areworking on LLMs. We're deploying LLMs. We're using OpenAI. They're using OpenAI
probably. Right now through my friends, I know, and they are very much ahead
with the technology. Okay, so then they won't be using when I was there at that
time we were using, generative model at that time.

[00:06:49] Nitin Singh: We were generating what do youcall that? So when you talk to a client, right? Client has some aim to achieve,
for example, let's say you're a client and you're from FFG industry, you would
say, okay, my [00:07:00] aim is to, my motivemission is to actually increase my revenue by X, percent in supply chain in
this region, right?

[00:07:06] Nitin Singh: Yes. Now this text, we cangenerate based upon their activities. So we were doing in Gartner. We were
doing this text generation long back, even before this through LSTM encoder
decoder network. We're generating those checks and all at that time, so they
have the technical architecture.

[00:07:22] Nitin Singh: We have AWS access. We had databrick access, so they will provide you everything and you just show the result.
So that's the beauty. If you're motivated to do things, you will have all
access to all the things. Otherwise, in different companies, you will say we
don't have AWS access. Thank you. Okay, Google collab.

[00:07:40] Nitin Singh: It's not it's not confidentialdata. You can run on Google collab and see, okay, take a subscription of Google
collab for one month and run that it's not how it doesn't work like that.

[00:07:50] Andrew Liew: Wow. Now I can understand whyit's so fast because the data accessibility, the data custodian, data approval
is, and the data infrastructure, everything's [00:08:00]there.

[00:08:00] Andrew Liew: So it, it makes your work likejust, like that, just like magic, right? Okay. So like back then as a director
of data science, were you in charge of churning a lot of like reports or
services or were you in charge of figuring out new products? I don't know.
What, exactly do you do?

[00:08:16] Andrew Liew: It's

[00:08:16] Nitin Singh: a mix of those actually. But whenI talk about. Reporting and all. So another thing good companies actually
standardize this process. Ah, okay. So they're standardized. They have all the
Power bi reports everything in place pipeline working. So from that perspective,
I think they were sorted, so that's what when you need reporting you need to
streamline how it works You need to standardize the process once you do that.

[00:08:41] Nitin Singh: It's pretty straightforward.Otherwise in my previous companies, there could be issue A lot of business
people saying, okay, I want this matrix. I don't want that matrix. All those
things to do something you need to follow the framework like SCR situation,
challenges, resolutions. And the Gartner culture is also very much influenced
by [00:09:00] McKinsey as well.

[00:09:00] Nitin Singh: A lot

[00:09:01] Andrew Liew: of people from

[00:09:02] Nitin Singh: McKinsey, that's what I'm saying.That's company. Have a lot of, strength, which nobody would actually know if
they haven't worked in Gartner. So that's, what we follow the structure. So
once you start following the structure in a more structured way, you work.

[00:09:16] Nitin Singh: I think things become easier.Otherwise we'll keep struggling to keep just setting up. The basic analytics
would take a lot of time. Otherwise.

[00:09:23] Andrew Liew: It's definitely structure isgood, but how does Gartner constantly evolve to be technologically advanced,
data advanced, because structure it, helps, but over time, things change new
customer demand units where, do you think that is happening in that sense?

[00:09:38] Nitin Singh: I think it's actually about theleaders because leaders actually, Understand the importance of automation.
Okay. It's not data science. It's not RPA or anything. It's mainly about
automating. Okay. When you actually look at a leadership level. Yes. If you
believe you can actually automate XYZ work.

[00:09:55] Nitin Singh: You will do that. Data is just atool. Yes. RBA is just a tool, but you [00:10:00]need to look at the outcome return. So it's driven by the business. No, it's
driven by the leadership. So that's why I say it's leadership who actually
drive this initiative. So leadership understand importance of data sciences,
all these things, then they spend putting their own budget into those areas,
higher quality people, and then do the initiative at their own risk and then
show the result.

[00:10:22] Nitin Singh: So that's why I said thatleadership in Gartner is really brilliant, to be honest. So I agree with you

[00:10:28] Andrew Liew: leaders definitely influence theway how the organization moves, but like in large organization as a company
like Gartner, it's considered a decent sized company. Whenever they want to
advance something, they want to check, move to something new, they have to
change from something old, right?

[00:10:42] Andrew Liew: And change is, human beings don'tlike change, no matter what. Even a data scientist, maybe he's so good at, like
you say, learning linear regression, asking him to use deep learning or LLM or even
a new package. He will probably be like I I got to go my, my wife is calling
me, oh, I got a girlfriend.

[00:10:59] Andrew Liew: How, [00:11:00]does leaders at, Gartner enable that change management to constantly innovate?
From a data and

[00:11:06] Nitin Singh: you use a very good word. Keepinnovating. I think that's the motto there as well. It's not important that in
the first try you succeed. So this this, I let, I was talking about, it did not
come in the first go.

[00:11:18] Nitin Singh: You try different things, youfail, but you learn from them. And that's what at we learned that and it was
all done before I joined Gartner, to be honest. So I had a luxury of
methodologies. For example a lot of time what happens is you do initiative and
you hurry up that initial process of vetting it, the pilot, measuring the
efficiency and then ultimately you end up the whole year effort goes into the
waste because the output is not something with business wants.

[00:11:44] Nitin Singh: Oh yes. So the failure teaches,it's very much important to spend those time in the pilot. So we spend a lot of
time in the pilot to ensure what's. The change we are doing, what change we are
doing is actually accepted by business and business actually [00:12:00] understand it. And a lot of companies,they will just go to they will just talk to a few people, 2, 3, 4, 5 people and
say, okay, business is fine.

[00:12:08] Nitin Singh: Let's do it. It doesn't work likethat. So we scale that part as well. So what we used to do is we used to create
some model. Let's say we create a generative model. Yes. The LSTM. We will,
publicize that in a way that will create an application, mini application and
dockerize that application and share the link with all the business and all,
and we will provide them functionality to test it at their own time, provide
the feedback, and then we look at the results.

[00:12:32] Nitin Singh: So that way we were actuallycapturing a lot more to test what we want to do compared to just. So basically

[00:12:38] Andrew Liew: you, you make the cost of risksmaller and you. You permit knowledge and capability of white so people can
say, Hey, this is a good idea. It's easy to test. Let's do it. And then
eventually things picked up and this whole culture and you also learn from

[00:12:52] Andrew Liew: Is that how it

[00:12:53] Nitin Singh: works? Exactly. Keep learning.And you know what? So company definitely have a culture, but somebody. The [00:13:00] leader who comes down come, and join acompany. I think impact of that leader is very important. So we have, I think
at that time we had Ken Davis, managing the whole vertical and that guy was
brilliant, man.

[00:13:13] Nitin Singh: You look at. Talk to that person.You'll learn a lot. And there were Kurt Cohen, Ken Davis, Malini Vittal. These
people are brilliant guys. This, so Malini is one who is from IIT Madras. A
brilliant, lady. I still keep pinging her and all, but yeah. These are the
people you learn from.

[00:13:30] Nitin Singh: And the impact they have basedupon their strategy is brilliant. So they basically came up with the idea of
having a data science. No. So that's, what important. You need to have a good
leaders in place and they can drive the initiative, man. They can turn, the
whole situation. Yeah.

[00:13:46] Andrew Liew: So I'm just curious.

[00:13:47] Andrew Liew: So how do you eventually from youwere doing well and you learn a lot and things was doing very well at Gartner.
How do you, eventually move to the next chapter, which is the current chapter
of data science at Molladay.

[00:13:57] Nitin Singh: One of the, one of the problemwhich I have [00:14:00] is I want to run veryfast.

[00:14:01] Nitin Singh: A lot of times and I used tothink that I'm in I'm nearing my 40s. I'm 30, 38, 38 right now, 39.

[00:14:09] Andrew Liew: I'm already in my 40s.

[00:14:10] Nitin Singh: So I'll tell you what happenedactually. So I used to think my peak time is over. It's just in my early 30s or
maybe in the late twenties my But then I talked to one of my, I would say one
of my uncle he actually led he was in Genpact.

[00:14:24] Nitin Singh: He led the whole Europe areas. Sohe's a COO in one of the pharmaceutical companies. So he told me, see this in
my life, this very small interval where people have actually, tell me
something. I catch on to that. I keep that in my heart. What was the timeless
advice? Please

[00:14:40] Andrew Liew: share.

[00:14:40] Nitin Singh: Yeah. It was that it's never toolate for anyone. Okay. Even at 60 year old or 55 year old, you can still do
things which you want to do in thirties or late twenties. So don't think of age
as a parameter, which defines your passion and success. Age is not a criteria and
that's what happened.

[00:14:56] Nitin Singh: That's what I realized later onin my career a lot of people are knowing us. [00:15:00]50 year old, they're still coding and believe me, they are a better coder than
you can ever know. They know everything. Okay. They're not like a makeshift
coder or probably been coding for 10 years. They want to become a manager.

[00:15:10] Nitin Singh: Then it's not like that. Theyreally are good. They know they're what they're doing, but that's what I
learned. That is what I realized. I just, I need to keep evolving, keep
discovering, keep trying to ha the ones who become complacent once you become
complacent, then you won't be able to grow.

[00:15:25] Nitin Singh: If I had become complacent inGartner probably, I would not have learned new things. And that is where the
change comes. So I never planned to leave Gartner to be honest. Was it because

[00:15:34] Andrew Liew: bumped into some friends atMuller Dean and they said, hey, come and join us or what? What's their story
about? So

[00:15:39] Nitin Singh: like I told you about one of the,okay.

[00:15:41] Nitin Singh: So, the guy who I talked to is wehad a common friend. Okay. The common friend is the same guy who is heading
investment in Vietnam one of the P firm. So we are very good friends. So I was
he's only one of the very few guys I'm still in contact after my college. So I
was talking to him and he, told me that [00:16:00]he, has one friend Who is based out of Singapore, looking for someone in data

[00:16:05] Nitin Singh: Now, I already had a Googleprocess going on before that. Yeah. So I cleared that, but I realized Google is
not something I want because after 15 years, I realized the brand name is
nothing. It's, more about what you learn there that's more important. So the
time passed for me. So I did not, I opted Molodin.

[00:16:21] Nitin Singh: And the person I was talking to,he was also one of a kind man, a brilliant guy. He's he's handling he's the
head of strategy here. I think it's Ankit. Oh God. I don't think I can even
manage his, the way he manages work. He's always working man and brilliant guy
very talented, folks.

[00:16:38] Nitin Singh: So I talked to him, I explainedhim what I did and we did the brainstorming. I think that was it. I really
liked the culture of this company because they approached me as well. It was
very warm. They were listening to me. They were not really bothered about Hey,
how much is going to ask? They were really actually very good people.

[00:16:56] Nitin Singh: So I, I came to Singapore it'sbeen around one [00:17:00] year. Wow.Brilliant. The team is brilliant, and the way they work. It's the actual
startup thing. I realize after coming here the way strategy changes and how
it's implemented, still the startup are very nimble in the way they were.

[00:17:13] Nitin Singh: You

[00:17:14] Andrew Liew: give the audience an example ofwhat is nimble relative to Gartner or relative other companies. You've

[00:17:18] Nitin Singh: got example, so let's say in abig company You have to change your strategy. For example, let's say you're
focusing on, you have these much you're forcing focusing on customer base aid.

[00:17:29] Nitin Singh: Okay. And let's say you realizebased upon the research that you need to move to customer base just change it a
bit, change the strategy a bit, change your marketing strategy and all those
things. Okay. Now that project would take a quarter in a big company. Wow. A

[00:17:43] Andrew Liew: Okay.

[00:17:44] Nitin Singh: Yeah. Because see, when you'redone, see.

[00:17:47] Nitin Singh: Strategy would be communicatedthrough technology and technology, business requirements, system design,
development a lot of things. It takes time and testing, regression testing all
these things. These things I learned from [00:18:00]Nagaro, all the how you manage a project.

[00:18:02] Andrew Liew: And then what happened in astartup, like what's the cadence, like how fast that was three months here is.

[00:18:07] Nitin Singh: They're very fast, man. Thisnight you're talking something and the next morning you have the feedback from
the ground that I think, wow, it's this morning,

[00:18:16] Andrew Liew: right? It's like McDonald's. Imake a McDonald's burger.

[00:18:19] Nitin Singh: I would say the faster than that,man. He's a brilliant folks. So that's what I said.

[00:18:22] Nitin Singh: The leadership is very important.So we have Andrew Who who is our c e o. We have Andrew Tan, who's a C F O, and
we have Anket, all these three. When they talk when they talk, lot of times you
feel, man I need to catch up with, they're very fast. Okay. Because they have a
very in-depth knowledge, of the industry and the way they talk, the way they
come up with strategies, the the frameworks. That's that's something you always
want to be there so the more you see them talking, the more you learn,
actually, So that's what, how do you

[00:18:54] Andrew Liew: know that, the strategy works orit is executable?

[00:18:58] Andrew Liew: I'm just curious because you saythat they [00:19:00] have so many strategiesand

[00:19:01] Nitin Singh: When, you talk about it, you alsohave people from, so text from data science perspective, if it's a
brainstorming going on, if I'm there, I can give my inputs that this is
possible. This, so we capture low hanging fruits.

[00:19:12] Nitin Singh: Everybody gets low hanging fruitsto start with. Focus on low hanging fruits and lot of time a little higher
hanging fruits. Also, we focus on so it's it's all team effort when the team
decide, OK, this can be done. And everybody extend, in a startup, everybody extend

[00:19:28] Nitin Singh: They try to, do things, try toprove they are worthy of being there because everybody's so competitive. And
that's how a team should be. Teams should be motivated in a startup,
especially. And one, one, another thing which so there was a scenario in the
company. A situation probably, I think, five, six months back.

[00:19:46] Nitin Singh: Yes, certain report came, certainoutcome came. So Andrew was looking at the Andrew who he was looking at the
output and I cannot tell you what exactly that was, but I'll tell you the gist.
People would look at things. OK, man, it's not going well. It's not [00:20:00] doing well your strategy. What I learnedfrom him is that he looked at that and figured out, I think we actually are doing
well, if you look at this area, kind of penetration we have, nobody has that,
if we can reduce the risk here, we can have more returns, looking at data from
a different perspective, from a little positive perspective, and try to come up
with a strategy to improve it further, rather than saying it's not looking
good, man, we need to come up with a different strategy, probably.

[00:20:25] Nitin Singh: So these things are you, don'tlearn these things from normal big companies because change management is very
difficult. It takes time. I,

[00:20:33] Andrew Liew: I is it because in big companies,the people at the top, they are used to sew. The, speed of information and
decision making, it's slow because the many layers, by the time you reach to
them, like things would've changed.

[00:20:45] Andrew Liew: Whereas in a startup the, layeris so thin. You implement, you see it works, you implement Seamless. Is that,
the case or how do you explain that?

[00:20:52] Nitin Singh: So yeah, hierarchy definitely whywe have hierarchy is to reduce risk as simple as that you reduce risk and even
in [00:21:00] startup, when we implementstrategies, we reduce risk as well, but before reducing risk, we need to
understand the risk as well, right?

[00:21:07] Nitin Singh: A lot of time you do a pilot andunderstand, okay, what's the limit and how much we are exposed if everything
goes down so you have that idea in your mind. So that is the difference. So
it's much quicker in a startup. And not every startup can be that fast. I'll
tell you this thing you know you need a.

[00:21:23] Nitin Singh: You need a good motivated team,passionate team to make it a success. It's not very easy that you know in a
startup if everything is ready. Can be like very fast. No, it doesn't happen like
that. And also in a startup, the the final decision, right? Not anybody can
take a final decision. So you need good leaders.

[00:21:43] Nitin Singh: So again, everything boils downto being a good leader. So even I aspire to be a good leader in future and I
can never be that leader, but I still keep doing that. Keep learning that. So
that's how I motivate myself and people should do that as well. They should not
think that they'll reach that level and it would be like fine.

[00:21:59] Nitin Singh: [00:22:00]It's always improving.