The value proposition of machine learning in FinTech is manifold, thanks to technologies that reach underserved markets and address their needs !

Pratyush Patodia, our next pathbreaker, leads a Data Science practice at a startup based out of Florida, that works across all industries to engineer and derive insights from data through statistical modelling.

Pratyush talks to  Shyam Krishnamurthy from The Interview Portal about his work on a machine learning model for instant approval of online loans for new-to-credit customers who are typically ignored by banks.

For students, Machine Learning is supposed to eventually benefit society by complementing human skills and allowing us to focus on lateral thinking !

Pratyush, tell us about your early years?

I grew up in Bombay, and studied Science in my +2, which was based on the advice of my parents. Though I had some thoughts about studying Economics, they explained that it would be easier to pivot to other fields from Science, which is very good advice even today.

I was not always good at academics. But the influence of a few good teachers, especially when I was in 7th and 8th grade allowed me to do significantly better. Interestingly, my interest in subjects would be highly correlated to the competence and enthusiasm of the teacher, so History was my favourite subject for one short semester!

When preparing for the 12th grade mathematics exam, since I was one of the few determined students not to go down the engineering or medical route, I decided to attempt the optional section of Statistics instead of Geometry. I found that I enjoyed it and was good at it. This was the start of my love for the subject.

My father is a chemical engineer from IIT Bombay, my mother is an MBA from ICFAI, and my sister is a Masters in Science (Maths) from UChicago.

What did you do for graduation/post graduation?

Given that I had an interest in Economics, and discovered Statistics later, I wanted to do my graduation and post-graduation in these subjects. 

St. Stephens was considered one of the best colleges in the country. I was selected for the interview, but did not perform well. So I settled for St. Xavier’s (Bombay), which in hindsight I am very grateful for, as I made some great friends there.

I knew I wanted to do a post-graduation since the job market was still recovering from the financial crisis of 2008. I decided this late in my final year of graduation. ISI Kolkata was considered the best, along with IGIDR and Delhi School of Economics. I got selected for ISI as well as IGIDR, chose ISI, and did not even give the test for DSE (Delhi School of Economics).

Tell us, what were the drivers that made you choose such an offbeat, unconventional and unusual career?

My parents, especially my father, had a good vision and knew that statistics would become pivotal in the future. My math and economics teacher from 8th and 9th grade was also influential in shaping my interest in economics and math.

My math teacher in 12th played an important role in exposing me to key statistical concepts. The statistics department at St. Xavier’s was instrumental in laying a very strong foundation for my conceptual understanding, something I rely on even today on a semi-regular basis.

Doing my post-graduation at a rigorous institute like ISI further honed my skills. Starting my career at EY gave me a problem solving bent of mind, and exposed me very early on to hands-on machine learning experiences. My role afterwards at ABFL (Aditya Birla Finance Limited) took me on a steep learning curve from a technology standpoint

How did you plan the steps to get into the career you wanted? Or how did you make a transition to a new career? Tell us about your career path

I had two internships, one at a boutique investment banking company, and one at BSE. The BSE internship, although short, ended up making a difference when I got placed at EY from ISI.

With my internship at Singhi Advisors, there was not too much scope to use my statistical or economic knowledge directly. It was more about understanding the context of the deals being pursued and supporting said deal-making efforts with the required research on the companies, industries etc.

The BSE internship had a much more direct relationship with the knowledge of statistics. My supervisor wanted me to research a concept called copulas which could be used to update the current methodology being used to determine the amount of margins required from transacting parties.

My next stint at EY was in consulting, where I developed and validated machine learning models for financial institutions across the North America, Middle East and Asia markets. These were mostly models that helped clients determine if their capitalisation and provision levels were adequate, especially in the face of potential stress situations such as the 2008 financial crisis. Examples of the some of the model types developed were linear regression, logistic regression, survival analysis, and decision trees.

At ABFL, the role was similar, but I was developing machine learning models for Fintech business use cases rather than just regulatory purposes. Additionally, since these models had to be deployed, I had to upskill myself significantly in technology.

At ABFL, the value proposition of the fintech arm was threefold: convenience (get a decision on your loan near-instantly), scale (being able to maintain the integrity of processes, such as underwriting), and target market (we wanted to be able to capture the market that was typically ignored by the banks, such as new-to-credit customers). Developing machine learning models was critical to power all three of these objectives, the fact that these models were deployed as micro-services meant real-time approval / denial of applications. It also meant that as the business scaled, all we had to do was scale the size of the micro services (which is much easier and cheaper than, say, scaling a team of human underwriters). Finally, the bespoke models developed were able to consume information beyond just credit history to assess the risk for unserved segments such as new-to-credit.

Currently, I am leading a Data Science practice at a startup based out of Florida.

One piece of advice that was often given to me but I did not always heed it (believing that meritocracy would find its way instead) is to network. You never know what opportunities can arise.

How did you get your first break? 

Campus placement

What were some of the challenges you faced? How did you address them?

Challenge 1: Transition from theoretical understanding to practical wisdom took a few months after college. The guidance of seniors in this regard helps immensely

Challenge 2: Learning to set appropriate expectations for yourself based on self-awareness is important so that you can grow at a steady pace. I think it takes at least 3 years of experience for maturity (or at least that was the case with me)

Challenge 3: Soft skills are critical, particularly people related. It is important especially when you start getting managerial responsibilities to relate to your team members at a human level, even as you help them reach the pinnacle of their potential

Where do you work now? 

Currently, I am leading a Data Science practice at a startup based out of Florida. I am based out of Toronto and a large part of the team is based out of India

What problems do you solve?

We deliver data and technology based solutions that help businesses stay ahead of the curve.

We describe ourselves as a tech services firm with a product mindset. We work across all industries and problem statement types. An example would be a grocery supermarket chain for whom, as part of a pilot, we engineered and derived insights from data generated by IoT based RADAR devices in the store which track customer and inventory behaviour on an anonymous basis.

What skills are needed in your role? How did you acquire the skills?

Technically, statistics and computer science are the main skills needed. I acquired it from my education and professional experience. Non-technically, one should possess a problem solving bent of mind

What’s a typical day like?

On a typical day, I am guiding my team members to help them with their technical tasks, and setting expectations in terms of plans for the medium and long term such as product development, training etc. Additionally, I am interacting with clients to understand their pressing problems, proposing solutions, and guiding project execution to success. Finally I am interacting with the CEO to understand our strategy and align the broader objectives of the team to said strategy

What is it you love about this job? 

I love thinking in terms of data and technology and more broadly, in terms of systematic and structured problem solving. There’s something new to learn everyday!

How does your work benefit society?

Machine Learning is supposed to eventually help human society focus on more fruitful aspects of life, so that’s one way to think about it, although I am not sure I believe that entirely. I think people who have a problem solving bent of mind, are good with data, curious about technology, and want to make a big impact on the world should consider a career in Data Science.

Tell us an example of a specific memorable work you did that is very close to you!

A machine learning model for approval of online loans that I developed was deployed in production and was the first of its kind in Aditya Birla. It was a proud moment for me and I was even awarded for it!

Your advice to students based on your experience?

Balance is critical. First take care of your physical health and make sure you have enough to survive. Then take care of your mental health so that you are self-aware and don’t burn out. Then take care of your relationships (family, friends etc.). Then take care of your societal / community obligations. Finally, think about your spiritual journey.

Future Plans?

Maybe a teaching role in some capacity at some point.