Headline

Machine Learning Lead | PhD in Statistics | Big Data + AWS

About

At JPMorgan Chase, our team leverages advanced AI and ML techniques to develop innovative solutions that drive the financial industry forward. My role as a Senior Associate centers on creating robust analytical models that inform critical decision-making processes and mitigate risk for the bank. With a strong foundation in quantitative research, I've contributed significantly to model development at Wells Fargo, enhancing predictive capabilities and strategic insights. My commitment lies in harnessing data to uncover opportunities that shape the future of finance.

Location

Bengaluru

AI stack

anthropic apis

antigravity

claude code

codex

cursor

ChatGPT

Work experience

Vice President of Artificial Intelligence at JPMorganChase, Sep 2022 - Present · 4 yrs, On-site, • Architected an enterprise income prediction model used across Credit Card, Auto Loan, and Business Banking; combined XGBoost regression with a custom loss and conformalized quantile regression to produce reliable 90% prediction intervals and improve DTI-based risk separation by 12%. • Built and operationalized an early delinquency detection model on AWS to distinguish transient/sloppy payers from high-risk accounts by modeling transitions from early delinquency to charge-off; used XGBoost with Boruta-SHAP feature selection and HyperOpt tuning, capturing $50M incremental charge-off mitigation versus the prior strategy. • Developed graph-derived features using TigerGraph to improve prediction of repayment ability and early debt-settlement propensity; integrated graph signals into a regulated model development workflow. • Led a comparative study of sequence-based approaches (1D-CNN, LSTM, attention) to detect non-monetary account takeover (ATO) events (unauthorized account changes) as precursors to monetary fraud; informed a proactive friction strategy to reduce downstream loss while limiting customer disruption. • Strengthened interpretability and governance by developing global/local explainability frameworks and model documentation for account management and collections use-cases; mentored associates and interns via structured ML training.

Senior Quantitative Analytics Specialist at Wells Fargo, Oct 2021 - May 2022 · 8 mos, Charlotte, North Carolina, United States

Quantitative Analytics Specialist at Wells Fargo, Jun 2019 - Sep 2021 · 2 yrs 4 mos, Charlotte, North Carolina Area

Graduate Teaching Assistant at Michigan State University, Sep 2018 - May 2019 · 9 mos, East Lansing, Michigan, 1. STT 201 - Statistical Methods 2. STT 861/862/867/872 - Graduate Statistics Courses

Data science intern at Liberty Mutual Insurance, Jun 2018 - Aug 2018 · 3 mos, Boston, Massachusetts, 1. Built a new Generalized Linear Model with Group Lasso Penalty, gradient boosting machine, boosted trees to classify Business Insurance customers. Modeling started with data cleaning, binning variables and with selection of important variables. Model validation with ROC, sensitivity and Lift chart. Software used SAS & R. 2. Application of Gradient Boosting Machine for variable importance and variable selection. Propensity score matching to identify causal variables. Helped business to identify important factors for customers sensitivity towards Quoted Premium. Statistical tools used R. 3. Developed a Bayesian Structural Time Series model in R to forecast future Business Quotes conversions by identifying important regressors. BSTS model obtained more accurate forecasting results than existing ARIMA model with potential profit to Business.

Data Science Intern at The Walt Disney Company, Jan 2016 - Jun 2016 · 6 mos, Orlando, Florida Area, 1. Developed model based clustering with mixture gaussian distributions to cluster Disney products based on their environmental hazardness. Worked in this short 2 months project single-handedly with the guidance of a Post-doctoral statistician S. McKay Curtis. Statistical tool R. 2. Built a generalized mixed effect model in SAS to predict future demands of Disney Resorts, which improves prediction accuracy of pricing tool to achieve revenue increment. 3. Proposed new modelling ideas to build the promotional tool such as incorporation of spline regression to handle time varying predictors, which had direct effect to improve prediction. 4. Trained incomplete and inaccurate data using SAS required for model. 5. Extensively studied price-demand relationship by working with different functional form of price as regression covariates.

Graduate Research Assistant at Michigan State University, Sep 2014 - May 2019 · 4 yrs 9 mos, East Lansing, 1. Building mixed effect models for Cancer Patients on Oral Oncolytic Agents using SAS. 2. Developed liner mixed models using the development data of African children in SAS.

Doctoral Student at Michigan State University, Aug 2014 - May 2019 · 4 yrs 10 mos, East Lansing, Michigan, 1. Dissertation topic: Bayesian Variable selection and Functional Data Analysis- Application to Brain Image 2. Bayesian Lasso models for classification with selection of high dimensional functional predictors. Application focuses on longitudinal MRI data from ADNI. Projects Undertaken: 1. Implemented a project on Machine Learning with application of Mixed Effect Neural Network model in structured MRI data to predict Alzheimer’s disease occurrence. The model performed much better in comparison to a Longitudinal Neural Network Model. Analysis done in R. 2. Deep Learning models with Feed-Forward Networks, CNNs, RNNs using PyTorch and TensorFlow in Python. 3. Bayesian Nonparametric model for Community Detection. The model and statistical methods in this project built on a well-known dataset of National Collegiate Athletic Association. We successfully detect the actual communities in the network using Chinese restaurant process and Dirichlet process.

Education

Doctor of Philosophy - PhD, Statistics, Michigan State University

Master of Science (M.Sc.), Statistics, University of Calcutta