I've built quantitative finance systems for stock prediction and hedging, including FinSenseAI, a full-stack ML platform that generates next-day return predictions across 20+ stocks.
For FinSenseAI, I engineered 80+ leakage-safe features from market data, news sentiment, and Google Trends, reducing prediction error by 20% with a Random Forest, Gradient Boosting, and XGBoost ensemble. I also deployed FastAPI REST APIs, built event-driven backtesting, and resolved production failures caused by NaN and Inf values from financial data APIs.
In my delta hedging simulator, I benchmarked Black-Scholes and deep-learning strategies across 10,000+ Monte Carlo market scenarios. My GRU-based deep hedging model reduced CVaR 5% tail risk by 42%, while adversarial market-path generation reduced it by 80% versus the Black-Scholes baseline.
As an undergraduate researcher at IIT Jodhpur, I developed end-to-end commodity price forecasting pipelines and ranked in the top 30% of a national Kaggle predictive modelling competition on real-world financial data.
