Abhijith Asok is a Seattle-based Data Scientist, data science educator and mentor who has been working in the field of data for nine years. Following a joint bachelor’s in engineering and a master’s in mathematics in his home country of India, he entered data science through a chance internship in data analytics.
Over the next few years, he gained significant experience in data science across healthcare, communications and social science, following it up with a second master’s degree in data science with a focus on healthcare, from Harvard University. Following that, he started working as a data scientist at Microsoft, where he continues to contribute today, as part of the data science team of the Core Operating Systems within Microsoft Azure – Microsoft’s Cloud Computing Platform.
Outside of his primary work with applied data science and AI, he also carries a passion for incorporating data into other aspects of life with significant practical impact. An advocate for gender equality, equity and inclusion, he calls out his time leading the data science team of Safecity, a non-profit that uses data as a tool to fight gender-based violence, as one of the most fruitful experiences of his career. As part of his time there, he has worked with researchers and helped mentor students from Stanford, UPenn etc.
He has multiple research papers and two pending patents to his credit, and fosters a deep interest in mentoring folks into data science and higher education in general and helps folks out with their questions on a daily basis. At present, he is in the process of teaching data science to a group of volunteers and through them, creating a structured educational data science program very focused on application-oriented learning, targeted specifically to bridge the barrier to entry to the field most folks without the prerequisites and experience face. In addition, he is also in the process of creating an organized and planned mentoring program in data science that he plans to launch very soon.
You can also learn more about Abhijith on LinkedIn here.
Here’s a quick summary of key takeaways:
During the conversation, Abhijith Asok, a data scientist and Harvard graduate, discussed his career journey and shared insights on pursuing one’s passion and navigating the application process. Abhijith initially pursued engineering and mathematics in college but discovered his passion for data science during an internship at a data analytics consultancy company. He emphasized the importance of an application-oriented approach and overcoming fear when pursuing a passion. Abhijith later pursued a master’s degree in Data Science at Harvard, highlighting the significance of a compelling statement of purpose and personalized letters of recommendation in the application process. He also shared his experience of joining Microsoft and the value of his educational background in approaching data science projects.
- He found that data science resonated with him and had an application-oriented approach that made complex math concepts more understandable and engaging.
- Abhijith emphasizes the importance of not feeling scared when pursuing something and the need to have curiosity and intrigue instead.
- He advises considering personal priorities when deciding on a career path, as passion should align with other important aspects of life.
- Abhijith’s decision to pursue a master’s degree in health data science at Harvard was motivated by the program’s focus on healthcare, which aligned with his previous experience in the field.
- Abhijith’s application stood out by presenting his journey as a story, acknowledging weaknesses, and demonstrating personal growth.
- He joined Microsoft through an AI program and went through a rigorous interview process that assessed technical skills and behavioral fit.
- When approaching complex problems and data sets, Abhijith advises starting with the data science methodology, which involves data collection, preparation, and understanding the problem before diving into modeling.
Listen to the podcast here, or find if wherever you get your podcasts:
Ryan Atkinson: Welcome, Avi. I’m super pumped to talk about your background and data science, data analytics, so thank you so much for joining us.
Abhijith Asok: Absolutely. It’s a pleasure to be here. Thank you, Ryan. Our first
Ryan Atkinson: question. I love asking these questions because you have a heck of a background. You’ve lived in India, London, Dubai, Boston, New York, Seattle. I mean, what has been your favorite city?
Abhijith Asok: Oh my gosh. I wanna say it’s none of those. It’s probably, it’s stanbul.
I’ve been to Istanbul only, like only I, I was in Istanbul for like five days. I wanna say five days. I went to, on this Turkey trip with my friend right after college for like eight days and five of those, five of those eight days we spent in Istanbul. And oh my God, it’s such a, such a gorgeous city.
Such a beautiful city. I have a thing for history. I particularly like history a lot and Istanbul. Or as it was called has such a rich history of, reli religion for one, but also war and conflict and everything too. And for me it represents like a [00:02:00] beautiful confluence of lot of things with like complicated history being part of it.
Interesting. And just being in that. City itself was like it. Just being in that space, being in that moment definitely made me feel like very, very good. For sure. And I would say that if, if money and the movement and the job and everything weren’t really a factor, I’d probably spend a lot of my time in Istanbul.
Ryan Atkinson: That’s awesome. How is the food in? Oh, I,
Abhijith Asok: I don’t even get me started. Oh my gosh. Oh God. They had this Turkey had this buttermilk sort of thing called Iran. I think it was the name, A y r a n. Personally, I’m a big fan of buttermilk. There’s like buttermilk in India. There’s like different kinds of buttermilk in the Middle East and everything in, in like Dubai, Saudi Arabia, whatever.
But they had this. Very light. It’s skimmed, I think. I’m not really sure, but very, very light buttermilk called Iran. Once I, once I started drinking, I couldn’t stop. I had to get like bottles after bottles because it was just so good. Turkish food is beautiful. I can just go to Stanbul again just for that.
Ah, I love
Ryan Atkinson: that. And as much as I can talk about travel, I can talk about travel for days. We are here to talk about some data science.
Abhijith Asok: Yeah, it’s a very good question. I would say that it was basically completely by chance. I went to college for engineering and mathematics and it was kind of like an integrated program where you spend five years to do both together, other than spend four and six separately for two separately.
So, I went to college for that. It was also kinda like a very socially driven decision because things might be different now because it’s been so long, but back then the Indian Society kind of had this like lean towards, engineering and medicine being like the only two like worthy things to do, quote unquote worthy things to do, which doesn’t make sense, but I know I was also like driven in that category. And I started doing engineering. Originally I wanna do English literature, but then it was like a [00:04:00] completely 1, 180 degree. Yeah. Away from that. And then I did, I studied engineering and I studied math, but neither of those explicitly like, spoke to me per se.
I wasn’t feeling connected to either of those. So, which also like prevented me from really actively like studying and gaining knowledge in them because I was , what’s even the point doesn’t make sense to me. So, and then I. Towards the end of college because it was this integrated program I had to do , like a final year internship, like a one year internship to finish the program itself.
And just by pure chance after a lot, lot of like, drama and everything, I ended up at this company called Mu Sigma, which is a data analytics consultancy company. And that’s where I did like a whole year of my internship and. Before landing there, I literally knew nothing about data analytics or data science.
Thank you. All I knew was like data entry where people look at numbers on a paper and put it into an Excel. That’s all I knew about it. But once I was there, I started understanding what data science is, what the possibilities of data science are what machine learning is, and How exactly, how exactly these things can help us in making the, like the world better.
It was comparatively a very new field at that point in time. But then I learned about its dif distinct components and mathematics, statistics and all of that. And most of all, I learned that. I have a passion towards it, that that kind of application oriented approach was working for me.
I discovered that I was good at it. It was coming to me. All of those complicated math equations that I studied in college, they suddenly started making sense because now I’m looking at it from like an application oriented perspective and which is, which is like a perspective that I carry forward to this day.
That, in order to like, learn things best, you need to have like an application oriented approach behind it. Right. And it just started coming, coming to me. I was becoming really good at it. I was very passionate about it, and I was like, okay. Finally I found something tech, technological and tech [00:06:00] related that I’m able to like do and let’s actually make like a career in this.
And that’s, that’s basically where it’s all started.
Ryan Atkinson: Interesting. It’s a little bit of a different career path than English literature, I will say. Yeah. Just
Abhijith Asok: a little bit. Just a little bit. I wanna
Ryan Atkinson: talk about like how you knew it was a passion though, because I feel like a lot of people, like at a school like myself, it’s clued.
It’s like, what am I really passionate about? Like, what do I like really care about doing? Mm-hmm. Like what are some signals to you that like you are passionate about something or what were those signals for you?
Abhijith Asok: I think the biggest one is you look at something and you don’t feel scared. I think that’s like a very big one.
As I said, in college when I was studying all those theoretical math, theoretical math equations and everything I just take one look at all those Greek symbols and integrals and differentials, and I, I’m like, oh my God. Like, what, what is even happening here? And I just like, give up. I just give up, close the book, and I go away.
Right. People I learned like to a whole year of like algebra when I was in, when I was in college, right? And they talk about like balls and a billion groups and all of these things. I, I, I’m like, okay, I get it. This is great. But, I, I’m, I’m not just not able to relate to any of those. And on, in the, in the con, on the contrary, in contrast, oh my God, what?
I can’t English in contrast. When I actually got, got over to the internship part of it when I saw a bunch of these same equations or similar equations for that matter calculus for the most part, I looked at it and when I was, since I was looking at it from the perspective of actual real data and I was able to like look at them from the perspective of actual real data.
I wasn’t scared at all. Like I didn’t feel like closing the book book or I didn’t feel like. Shutting down my laptop and going away. I actually wanted to know more if there was like a long equation that I didn’t understand, I wanted to spend half an hour to understand it. Whereas in the past, even if I understood like half of the equation, I wouldn’t spend five more minutes to understand it because I just wasn’t feeling it.
So I think when you look at something and if you’re not scared, or rather if your fear is sort of like [00:08:00] overcome by your curiosity and your intrigue, I think that’s like a very, very big signal that that is probably your passion. And then apart from that, I think you also need to look at priorities because it’s really easy to say, oh, follow your passion.
But at the end of the day, we live in a realistic world, right? And it comes with its own biases. I personally feel like. Every single person in this world is good at something or the other. That’s something I strongly, strongly believe. It’s just the unfortunate reality that not all of those are weighted equally the way the society is set up today.
So I think everybody’s definitely good at something. But if people gotta like, also think about their priorities. So for some people money is a big priority. Nothing wrong with that. If, if money’s what you want, that’s what you want. For some people, peace of life is what they want. Flexibility for some people.
Being closer to family for some people, right? So if your passion would sort of like take you away from a bunch of these priorities, that’s also something you definitely need to think about because I feel like passion, sometimes passion comes to you, but other times I definitely feel like you make your own passion.
Because if originally, like English literature was my passion, I still like to write, I still like to write poetry and, and creative writing in general I just fell into situations where that just wasn’t possible for whatever reason. And then, I discovered something along the way and sort of like made that my passion.
So I think idealistically I can be like, oh, if you’re not scared of it, just go for it. But realistically, there are all these other parameters as.
Ryan Atkinson: Interesting. So it, yeah, so it’s balancing like what, what are you not scared to do? Like what makes you super excited? It’s also a little bit of a balance of like, we have bills to pay and like we just be super realistic.
Not super realistic, but realistic enough where, yeah, those two, that is really like an area that you should explore if you are passionate, help pay
Abhijith Asok: the bills. Definitely. Yeah. It’s just kind of like when people apply for let’s say grad school, right? Yeah. They have all these considerations like, oh, how much is it gonna cost me?
Which city is it in? Is it near, near to one of the big cities or [00:10:00] in one of the big cities where I can easily get a job, et cetera, et cetera. And people, people don’t always like get all of, get to check all of those boxes. They usually have to give up on one or two to get the others right. I think it’s a similar way.
I think you just Figure out what your priorities are and see where your passion actually fits in there. And if you’re like, I don’t care about anything else, I only care about pursuing a passion. Fantastic. Go for it. It’s not the reality for a lot of other people and I want them to know that it’s totally okay to have other priorities as well.
Ryan Atkinson: I love that. And let’s talk about graduate school, cuz you went to a school called Harvard with your master’s in science and health Data science, which is so cool. You started that in 2017. Take us to that time. Let’s go back to 2017. Harvard, like Howard, Harvard, it.
Abhijith Asok: Oh my God. Okay, so I feel like we might have started in 2015 technically. So as I said in, in 2014, I went into, dropped into this new Sigma and I started learning data science and stuff a little bit more by the end of it. By my graduation time, I knew for sure that I really wanted to do something in data science, but I did not really have anything to show for it.
Like I only had my interest to show, I did not have any projects on anything to show because. Throughout my college I didn’t really do anything related to data science. Yeah. So it was mostly just my interest interest to show. And it was very difficult to find like mentors or people to like, mentor and like guide you through that process because No, I, I can understand from their perspective too, like, like why, how exactly they have limited time to, among all the other things.
And if they’re supposed to like tell somebody, they need to know that they’re serious about it. And there was nothing I could show. It’s kind of like a, like a cash 22, um, where I can’t, I can’t show you something without doing something. But then, if you don’t like, gimme that chance. I don’t have anything to show.
So basically for like the next couple of years or so I started saying yes to pretty much everything that came my way and started hunting for opportunities that I can do[00:12:00] rather than thinking about how I will organize my schedule and how I find time to do something. I just like said, yes, I’ll do it, and then figured out how I’m gonna do it.
, with the time and everything. I won’t advise anybody else to do that. I had to do that because of my situation, but that’s not an advice that I’ll give anybody for sure, because it takes like a big toll on your mental health and everything. Yeah. And along the process I started building my profile little by little.
I started having a lot of like diverse kind of experiences. But there still wasn’t somebody to sort of tell me whether I’m doing something wrong or whether something I’m doing can be done in a better way. In a field like data science, when technology keeps changing by the, by the minute or by the day, it just very, very fast changing field.
New, new things come up. A lot of the cutting edge technologies of today did not exist like two years ago. I mean, chat GD for example. Right? So, there wasn’t anybody to like show me that right track or tell me that, Hey, you might, you can sort of like do this in another way. And I was finding it hard to find proper mentors because of whatever I just told you.
Yeah. And that sort of like, Made me think about grad school specifically in data science. I already had a master’s, but I was like, it will definitely gonna, it’ll, it’ll definitely, definitely be beneficial to me if I am able to do a second master’s degree specific to the data science or analytics or something amazing.
And that’s sort of like start at this whole ball rolling. I think I, this made this decision in late 2015 and then I used like 2016 to do all the applications and gfl and whatnot. Um, About like Harvard, it was very as you said, the, the, the degree specifically in healthcare, oh, I shouldn’t say specifically in healthcare.
What they teach you is like very general things, is just that the examples and stuff that they use mostly from healthcare. I, it was not so much the brand name that sort of first attracted to me at fir attracted me to it. At first it was more like [00:14:00] Okay. First of all, when it comes to America, like in data science education, America was still definitely in, in the forefront at that, at that point in time, at least in 2016.
So it was like I really wanted to apply to America. Nowhere else that, that I kind of knew. But then I was looking at the list and I was like, the, the Ivy Leagues are supposed to be like really good with like quality education and everything. But where are they? I saw Columbia has like, data science program and I was like, where are the others?
Like there are like eight of them. Where are the others? So I started looking to each of them to see what, what was happening with each of them. A lot of them didn’t really like find anything that I could see. Mm. Um, and then I saw that Harvard was like, Launching their data science initiative. And there there are three programs underneath it.
And the first one was health data Science, and that was the one they were launching that year. The other ones hadn’t even like started yet. And I was like, among the different, uh, fields that I’ve been been in, in the past, especially specific to data science, healthcare was the one I had like most experience in because of Zs Associates, as you mentioned at the beginning, i’s a pharmaceutical consultancy.
So even though I was like, Officially like a business analyst. That was my designation there. I used to involve myself in a lot of like data science stuff in the company as well. And through them I learned a lot of, lot about the healthcare industry, especially in America. Yeah. And I felt like that sort of gave me like a really strong background to apply and that’s basically what made me apply.
Honestly, I didn’t even. Uh, I expect to get in. It was more of like, if I don’t apply, right? If I, if I knew about this and if I don’t apply right now, I’m gonna look back five years later and be like, what if I had applied? So I didn’t really want that, so I wanna just apply and I just applied it. I didn’t even like look at my application status for the next few months.
One fine day. An email came out nowhere and I’m like, oh, what is this? I’m like, oh, applied. They’re like,
Ryan Atkinson: Oh my God. I like how you said, um, I’m gonna apply, but I don’t really think I like, have a chance. I feel like so many people that [00:16:00] graduate from Harvard, like, or like mit, they all say the same thing. It’s like, I, I didn’t think I actually had a chance, but I applied anyways. I’m curious.
Abhijith Asok: I don’t think anybody ever, well, maybe there might be some people, but I don’t think most people ever expect to get into these spaces.
They hope. Right? Yeah. And that’s basically where it was. I, I didn’t even hope I was more, it was more like, I don’t wanna regret it later. Um, but look how things turned out.
Ryan Atkinson: Yeah. Wh where, where exactly were, you have to remember where you were when you got the Applic,
Abhijith Asok: right? Oh, yeah, definitely. Oh yeah. I was in Dubai.
I was my mom had, was living in Dubai at the time, so I was visiting her. Dad and I were visiting her and we were all in like the same room when that happened, right? And the email came, I was about to go to shower, go shower, and this, this email came in and I didn’t even like tell my parents that, oh, an email from Harvard came in.
I was like, okay, I’m gonna look. I’m gonna confirm, reject, and. Open the email I saw like congratulations, and I’m like, this can be right. And then I read it again and again. I read it like three times to make sure that I’m not missing anything. And then I tell, told my parents and being like the super sensitive person that I am, I just like started crying.
I started like crying in the room with everybody and they started crying with me. The three of us were just like crying for like half an hour after that. That’s very beautiful moment in my life for sure.
Ryan Atkinson: Yeah. That is so cool. And like, let’s talk about that application process in like 2016 specifically like for Harvard.
Mm-hmm. Can you take us like what that application process like looked like and was there anything specifically that you included where it was like, oh, that really made me stand out?
Abhijith Asok: Y I would say that much like anybody, everybody else the application want one of the usual requirements. Like Oh, like a personal statement or statement of purpose?
Yeah, like written letters of recommendation, resume, all of that stuff. But they also looked for a, what’s called a W E s evaluation. Okay. Which is basically for people who, dunno, that’s basically World Education Services. They’re like I think a private company that sort of does transcript conversions from one country to [00:18:00] the other.
So basically for for United States because Indian education system is very different from the US education system. They take your transcript and they do the conversion and whatnot. , that was also required from them. And I would say that what made me stand out. Something interesting that happened with the WS itself was that because my program was like an integrated program it, it had the structure of the program was like very different from like a traditional bachelor’s of engineering Yeah.
Or whatever program that people are usually used to. And I don’t think Ws really knew how to make that conversion effective. So when they made that conversion, they ultimately gave me like a converted GPA of like 2.84 on four. I was, that’s really not, that’s. Absolutely not accurate. Like, I know I don’t have like a stellar gpa, but 2.84 is like too low.
That’s, that’s definitely not right. So I reached out to WS and explained, Hey, here are the reasons, here is why it should be different. , they didn’t really. Change it at all. They’re like we won’t be changing it at our end. , this is final for us. And I just I was like, okay, what do I do now?
And then I decided to email Harvard directly emailed the admissions committee of the program and I was like, Hey, this is what happened. I can explain, here’s my explanation as to why. This transcript conversion is inaccurate. And they said that okay, they can’t really like forego the WS conversion, but what they can do is they can probably attach my entire explanation to the application.
And I hope they did that cuz I didn’t see my application materials or anything. They said that they’re gonna be doing that and I think that probably helped quite a bit because I was like, fighting for something, fighting for like an accurate representation of who I’m right.
Uh, I think that might have gone in my favor, I would say. Uh, but beyond that, I would say the most important. Components of the application and for any [00:20:00] application? I think in general, well, I shouldn’t generalize like that, but at least for a lot of applications out there the statement of purpose and the letters of recommendation, the combination of both of them, I think that’s what makes all of the difference a lot of the difference.
At least the the crispness of the statement of purpose is very important, but also it’s more important, in my opinion, for it to sort of like read like a story. Yeah. For a lot of people follow this approach of like, oh, I did this, I did that, and then I did this other thing, and look at me. I’m awesome.
How cool am I? I have no weaknesses, blah, blah, blah. A lot of people do follow that because they wanna like, portray them in like the best manner possible. They have, like, I have all the information, I have this word limit, how do I put everything in? Yeah. But then I feel like if it doesn’t read like a story it falls flat because people, even the.
Professors who are reading it, they might, they might get, like, you might get like two professors two minutes each spending on your application, right? Yeah. They have like four minutes to sort of like impress them, impress two people with your application. And for that, something as a standout.
And I feel like that kind of personal touch and the story based flow of your Statement of purpose. Like, oh, I did this, this is what I learned from it. If you failed, like why excited did you fail? And what excited did you learn from that failure? And how did that lead you into your next experience and whatnot?
And people, nobody likes to, nobody likes it when people sort of portray themselves as like a superhero, right? Yeah. Uh, you definitely, everybody has weaknesses. So being aware of your weaknesses and presenting them in a way, in like a positive way where you’re like, this is my weakness, but this is what I can get out from it.
Or just like the other side of that weakness where it. Is advantages in a way that you don’t expect? I think those are all things that are important, are important, and if your letters of recommendation can sort of like substantiate whatever you’ve written in your personal statement, especially with like, personal anecdotes and stuff, I think that makes a lot of difference.
So as [00:22:00] opposed to like a generic letter of recommendation from like a, like a super high level manager versus like a personal letter of recommendation from your immediate manager. I would totally go for the latter. Any point in time.
Ryan Atkinson: I like that. And let’s talk about another, uh, huge issue in your life.
So you graduate from Harvard in 2019. Mm-hmm. Shortly after you get a role with Microsoft, one of the largest market caps in the world. Everyone knows Microsoft. I don’t even need to hype that up. Um, I’m curious, like how did that job, like really, how did it catch your eye? How did you notice it? Um, and like, take us through like that application process.
Uh,
Abhijith Asok: so what happened again, just like Harvard, um, I wasn’t even planning to apply for Microsoft. I’m like, what’s the point? Yeah. And then the last semester, I think I was all set to graduate in May, 2019. Yeah. And in January, 2019, I was still. Thinking about like what kind of role I want and I’m still like, casually applying to positions, but like, not super seriously just yet.
Yeah. But I wasn’t even like gonna try for Microsoft or Google or anything of the sort, cause I’m, I’m like this is too much effort for like really no return. Right. Yeah. But then I had like an HR from Microsoft reach out to me on LinkedIn in January of 2019 and she introduced this sort of like program, like an AI program within Microsoft.
That they were in their second year or something. So it’s like a cohort based program, but it’s not like an internship. It’s like a full-time role. It’s more like an internal AI consultancy within, uh, within Microsoft, basically. Right. So the teams sort of does AI work for other teams that maybe either don’t have their own AI capabilities or they don’t really have, um, They don’t need the bandwidth or they don’t hardly have the resources for it or whatever.
So they basically hire people fresh out of like university or people within the early stages of their career with ai. Yeah. And [00:24:00] Sort of like trained them to sort of be like AI specialists in, in the company basically. And they, she introduced me to this program and I was like, alright, well if somebody reached out to me and if I don’t apply now, that’s gonna be like a, that’s not gonna be good.
Yeah. So I made an application and same way I, I never expected to hear back, so I forgot about it. And then I got like an interview email for like a phone interview I got on the phone interview. This is very quick. Like within a week or something. Yeah. Got on the phone, on the phone for an interview and they were very interested in my, the research side of things that I had done in my past and all of that as well.
Very deeply went into like the resume and then I again, I was like, all right. That went okay. But you know, that’s, I, I’m sure it’s like extremely competitive and I don’t really hear back. But then I heard back, and this was, this was pre Covid so the next round was like in person. So I went into the Microsoft office in Boston, the research center in Boston.
And I had like, I wanna say four interviews, 45 minutes each without any breaks. So as soon as one’s done, you are taken to the, um, I don’t know, courtyard or quarter angle or whatever it is for the next person will be there waiting for you to like, take you to their, to the next room. So that’s basically how it works.
We’re three technical interviews, one behavioral interview, and then after that, two days later, I got another for final phone round with the director of the program as well. So I would say six, six rounds of interviews in total. And I took around like one to two months for the offer to come through.
And I had an offer in March of 2019, about like two months before I graduated.
Ryan Atkinson: Interesting. I’m curious cause you went from Harvard to Microsoft and like how much did you learn from Harvard that was like actually like applicable to your role at Microsoft?
Abhijith Asok: I would say for the interview whatever I learned with learned at Harvard, Harvard was extremely helpful.
Mm. I, I think of some of the courses that I learned at Harvard, I think that that is the reason [00:26:00] why I tracked a bunch of those interviews. Interesting. When it comes to like actual, Work, it’s hard to say because it’s very project dependent. Mm-hmm. Cause in that team, I, I was in that team for two years.
In that team, it’s all project dependent. You spend on, you spend about five months or so on one project, and then you shift like another project that can look completely different from the first project. So it totally depends on what the project demands. But I would definitely say that the understanding that I, understanding about techniques that I learned from my courses at Harvard was very useful in sort of, Helping me think the right way.
You know, when I look at a problem, to be able to think in terms of like a certain kind of solution, I think that is really helpful because, , you can be a data scientist even if you don’t know how. Algorithms and stuff work. You could probably, there are packages these days for everything.
You can just like code in Python. You can like go and call something in Python and you can get things done. You can still be a data scientist and I’m not saying there’s anything wrong with that. If that’s, that’s what you wanna do, that’s great. But I think that knowing the math behind the techniques and.
Techniques and methodologies and stuff, uh, or even if you don’t know the map, being knowing, having enough like background that you can quickly go read something up for like 10, 15 minutes and like understand what’s happening there. I think that really helps you be like a better data scientist for sure.
Makes you think about things the right way. If you wanna, like, if you want to understand why your model is behaving a certain way, what it takes to like improve it that kinda understanding the math behind it sort of helps a, helps you go a long way towards making improvements and being more efficient and everything.
So in that sense, I would say that whatever I learned at Harvard was very, very helpful. Interesting.
Ryan Atkinson: I’d be curious, do you have any like frameworks that you use if you get like a complex problem, you have a complex data set? Mm-hmm. Um, is there like a framework that you use to like start simplifying this and like going after it?
Abhijith Asok: Generally the data science [00:28:00] methodology, and this is something that a lot of people getting fresh, getting into data science don’t really realize. We hear about ai, we hear about like machine learning. We hear about all these like fancy models. You don’t start doing that from day one. In fact, you probably won’t even start doing that at like 50% of the project because majority of the time in the project you actually end up.
Collecting the data, like preparing the data, making sure the data is accurate, putting it into the form that you want, et cetera. There’s a lot of steps coming in before you get into that part of it. So I think that going in, you should definitely forget about like the models and stuff you want to use if you have like a problem.
I think one of the most, first and foremost things to do is to like, Convert that into like a data science problem, mm-hmm. Uh, like for example, if somebody tells you like, Hey, I wanna predict the weather, and if that is kind of like the only sentence you get, right? And let’s say, let’s say this is like the CEO of your, of your company who doesn’t really know much of like data science themselves, right?
And the CEO comes and tells you, uh, she comes and tells you that, oh, you know what? Let’s, let’s predict the weather. That’s it. She walks away, right? What do you do? You can’t just be like, oh, I’m just gonna like go in and apply a gradient boosting model. I’m just gonna extract data from the internet, apply a gradient boosting model on it, and like I’m done.
No, it doesn’t really work like that. What exactly does it take to predict the weather and what kind of features do you need? What does it mean to predict the weather? Is it gonna be a time series model or is it gonna be like a tabular form of data? What gonna be Once you start making your, where do you get that data from?
Is that data even available? If not, are there like other kinds of data that you can use as a proxy for it or not? How clean is the data? How much time you do you need to spend cleaning it, forming it, pre-processing, and all of that. And then even when you start thinking about the. Start thinking about actual modeling and actual creating, like a data science model, machine learning model, or whatever you still think.
Start thinking about baselines, how [00:30:00] excited, do you know how good your machine learning model is? Like if I predict the average of the past five days for the next day is that, Already good enough that I don’t really need a machine learning model. So you create like a baseline and then you create like a bunch of rule based model.
Rule based models are extremely fast compared to machine learning models. So if a rule based model can give you the kind of efficiency and the accuracy that you want, you need to go with that. And then you get to machine learning. Then you do all these different kinds of explorations. Then comes.
Inference, like how do you understand the result of the machine learning model? How do you present all of these things that you did over the past 2, 2, 3 months or so to somebody who has absolutely no data background? Right? So this is like a whole process. And I would say that on the whole, you spend only about like 20 to 30% actually doing the modeling or the prediction part of things.
Everything else is sort of spent in like data prep and data cleaning and everything, and going in, you gotta like be really you gotta like be aware that this is the kind of process that’s awaiting you. Yeah. And I feel like generating that. Thought process, generating that problem solving approach.
I think that is definitely, definitely the best best thing you can do if you are looking like get into getting, getting into the field of data.
Ryan Atkinson: Interesting. And I wanna hit on one point on that. We’re, we are winding down on time here, right? I just want to hit on one point that once you do get all the models, you get everything together, you cleaned everything, you have this beautiful model, you’re super excited to present it to someone that doesn’t know anything about data science.
How do you effectively
Abhijith Asok: do that? Visualizations are definitely important. So. One thing that my manager told me, and this is something that I’ve heard from other people before at, at Harvard from my professors as well, is to sort of like, never use a
Ryan Atkinson: pie chart for
Abhijith Asok: anything. Okay. If pie chart is like, people think about the basic kind of charts, people know, bar charts, line charts, pie charts, whatever.
I’ve gotten advice from multiple corners that never use a pie chart. It kind of makes sense to me too, especially when you have like a bunch of different categories to [00:32:00] represent. When the categories start getting to be around like 20%, 15%, 10% and everything, it really doesn’t tell you anything. You can’t really compare between like 10% and 7% in a pie chart, right?
Doesn’t really tell you much. So I would say that choosing the right visualization is definitely going to be important and storytelling is gonna be important, ru So rather than be like, Hey, here’s, here’s like visualization of how we, , cut by salary and here’s the visualization of how we cut by geography, blah, blah, blah.
It doesn’t really like work like that. People need to be able to connect, especially people who are like higher up. We only have a lot of time. They’re looking for like actionable insights. You gotta be like, okay, here, this was the problem. This is why we, this is why we had to find a solution to this problem.
And these are the different considerations that we had to make. And this is the approach that we chose because of this reason. These are the things that. We learn from it and we can this particular insight that we learned sort of like feeds directly into this particular, I don’t know this particular consideration that we have in our annual report last year.
So this directly feeds in there talk in terms of things that they can easily understand as like a whole well packaged story. And I think that’s how I would do it.
Ryan Atkinson: Awesome. Well, Avi, last question for you. This has been awesome. But just general career advice. You’re obviously well accomplished. You’ve done a whole lot, you’ve done well in your career.
I mean, just what advice do you have to anybody, not even people that are just data science focused? Let’s actually just stick with data science focus. What advice is that people that are wanting to get into data science? My
Abhijith Asok: number one advice is always take care of your mental health. That is something that people discount left and right, honestly.
It’s very extremely important you can like kill yourself with all of the things that you can do. And, but then if you don’t really have like a good enough mental health none of that is gonna like pan out because you won’t be able to like, enjoy the life that you built for yourself. Um, and, and I promise you that if you really take care of your mental health, you’ll obviously feel content and happy.
In life in [00:34:00] general, obviously like nobody’s like a hundred percent happy. Everybody’s always dealing with insecurities and biases and all of that. But having that awareness about those problems is gonna be like really important. And once you, once you already have like a stable ground to stand on and once you have awareness about the things that you need to work on, that gives you like a really good route roadmap or like a roadmap to actually like work on them.
And having like a good mental health will make you like sit in front of the. Work you need to be done every morning with like a smile on your face or like, like have joy inside your mind. Uh, because even if what you’re doing is your passion, um, if you’re not able to like focus on it because of all these other things that are going on.
That’s just not gonna, not gonna pan out. But at the same time, I also wanna acknowledge that that’s also a very idealistic approach. A lot of people are going through a lot when, when you’re looking to like, survive, mental health is not really like a really high priority for you in those situations.
Uh, I wanna say like, be kind to yourself and there are definitely people out there who are willing to like, talk to you and like help you out. I personally, I’m actually looking forward currently in the process of like creating a data science educational program, very application oriented, because there is this.
Barrier to entry to data science, especially for people who don’t have that kind of background. Yeah, everybody looks for like math background and this requirement and task requirement and everything. For people who haven’t had the chance to study those, there’s this barrier to entry into data science.
So currently I’m putting together this program, educational program targeted at like those kinds of people where they can learn data science from scratch with like zero background. Yeah. In like very application oriented manner. And I’m also gonna be, Thinking about launching like a like mentorship program to sort of handhold people into data science and stuff as well.
That’s amazing. So there are definitely people out there willing to talk to you if you want that help. But most importantly, like definitely be kind to yourself. Awesome.
Ryan Atkinson: Well Avi, this was amazing podcast. Thank you so, so much for joining. And yes, thank you [00:36:00] so much
Abhijith Asok: for being here. Yeah, thanks for having me.
Appreciate it.