Mumbai, Maharashtra, India
12+ years of experience primarily in the domain of Financial Risk Management (Fraud Risk, Consumer Credit Risk and Scorecard Development), currently responsible for leveraging sophisticated Machine Learning Algorithms in a Big Data platform (usage of Apache PySpark/ Scala and Hive; usage of GPU for model training) for mitigating Fraud in Chase Consumer Banking Products like Credit Cards, Digital Money Transfers and Cheques with effective Model Risk Management as per SR 11-7 Guidelines. Have hands on experience in Predictive Modeling and Machine Learning algorithms (Linear Regression, Logistic Regression, CART, Random Forest, Gradient Boosting and XGBoost) and well versed in tools like Python, SAS, SAS E-Miner, Microsoft Excel and Query Languages like Teradata.
- Developed a decision tree using stochastic gradient boosting algorithm for fraud applications which qualify for Early MoB (month on books). It helps identify imposters who take-over accounts using genuine customer information. - Developed a Bustout Fraud model using Logistic regression to rank incoming credit card applications based on their propensity for fraud. - Developed models using gradient Boosting and logistic regression, to mitigate fraud in Chase digital products such as Automated Clearing House, Bill Pay, Quick Pay and Wire Transactions. - Extensively used of Machine Learning for Fraud Detection - Knowledge of Stochastic Gradient Boosting, Random Forest, Python, Angoss
- Behavior Scorecard for one of the largest US Bank: Developed scorecards (Early and Late Cycle) used for making decisions such as Credit Line Management, Collection queuing (priority calling, emails and payment authorization) etc - Check Book Forecasting – Built models to forecast Check Book utilization to minimize storage cost and maximize utilization - Knock-Off Score - Developed scoring model to predict the Bankruptcy behavior of the customer, at the time of origination
- Acquisition Scorecard for one of the Largest US Bank: Developed scorecard to acquire new customers for the Canada Personal Loans portfolio. This was a re-development of the application scorecard, due to change in underlying population and deteriorating performance of the score. The re-developed score led to better quality of acquired customers, improved the risk profile of the portfolio and reduced losses. - Exposure to validation of the various scorecards for US/Canada across various portfolios
- Conversion Model for US online retailer: Developed a propensity score of a member to activate in men/women division through a particular brand group. The model helped identify the drivers of a user’s purchase in a brand group, which in turn increased the activation rate and sales volume - Cost Modeling for US Pharma Company: Analyzed the prevalence /incidence rate and economic burden of a disease. Model was used to identify the cost effectiveness of the treatment/drug. Analysis was useful for the drug company, Insurance Company and the Government.
Data related work and to calculate sector specific grants