Even before moving to New York City to study at Columbia, Data Science Institute alumna Ritu Tak worked on the data science team of Axis Bank in India. She used machine learning models to predict the behavior of the bank’s corporate customers, especially which of them would become delinquent. This work helped her realize the power and scope of data science, so she chose to expand her expertise in advanced techniques. Here, Tak discusses her current role in machine learning for JPMorgan Chase and how the DSI master’s degree program helped her land the job. 

Tell us about your role with JP Morgan Chase.  

I am a quantitative researcher and machine learning associate in the Wholesale Credit Rating Group for JPMorgan Chase. My work mainly focuses on analyzing, redeveloping, and streamlining the internal rating process of the bank’s non-retail clients. Business needs and regulatory requirements mandate that banks rate internal clients according to their credit quality, which reflects their probability of default. These ratings are then used across lines of business, including commercial banking, asset- wealth management, and corporate and investment banking, for pricing, risk management, capital calculations, and loss estimations. The automation of the rating process using machine learning algorithms is considered strategic internally. And since I am the only team member explicitly trained in machine learning, I add a unique perspective to solving problems. I got a summer internship with JPMorgan Chase through DSI’s career service office and received the full-time offer after the internship.

Which data science skills do you use for your job?

I develop and maintain sophisticated mathematical models and modeling methodologies. My work often requires referencing the literature on classification models as well as experimenting with new techniques. This requires that I code algorithms from scratch, since ready-made Python packages are not available for newer algorithms. Creating these models requires an innate understanding of concepts like cost functions, optimization methods, and estimation methods. JPMC is one of the biggest banks in the United States, and inefficient programming and processing can lead to erroneous processes that can run for several days. The use of technologies such as Linux boxes to access high memory and compute power, knowledge of bash programming, parallelization of tasks in Python, as well as the use of memory-efficient data formats is all part of my job.

What did you study at university in India?

I have both a bachelor’s and master’s degree in physics from the Indian Institute of Technology, Kharagpur. The physics program offered subjects varying from the study of quarks to evaluating the dynamics of DNA. The major focus of coursework, however, focused on the mathematical concepts required to understand and handle any physical phenomenon. While pursuing these courses, I realized my interest lay more in the statistical and programming techniques, rather than the specific physical systems themselves.

How did the DSI master’s degree program prepare you for your current job?

The course Algorithms for Data Science was vital in helping me learn to write efficient scripts and in calculating the run times of algorithms. In my daily work, I also use estimation methods and hypothesis testing that were covered in the course Statistical Inference, as well as optimization methods that we studied in the machine learning course. Additionally, the Applied Machine Learning course provided me with much required finesse in using scikit-learn for supervised and unsupervised tasks. The assignments for the machine learning and applied machine learning courses also helped me understand and apply core data science concepts.

Do you have any advice for current and prospective data science students?

First, focus on building basic machine learning concepts from the ground up. Next, I would advise them to work on their coding skills, using portals like HackerRank and LeetCode. Both machine learning and coding are very beneficial when it comes to cracking job interviews.

— Robert Florida