I built and run RiskSignal, a live portfolio-risk analytics product for Indian financial advisors. It parses broker exports and scores portfolio risk, then delivers PDF reports by email.
For RiskSignal, I wrote a regression- and mutation-tested suite of 600+ automated tests, fixed asset-type misclassification, and set up production monitoring with Sentry and scheduled-job alerts.
In my NIFTY 50 volatility thesis, I tested 20+ models and found that HAR-RV outperformed the deep-learning models for volatility prediction. A chronological per-stock split helped prevent data leakage.
As a Data Science Intern at Spartificial Pvt. Ltd., I built and deployed a deep-learning model for real-time exoplanet prediction with 90% accuracy. At Accion Labs, I developed an automated loan-approval recommendation system using PostgreSQL.
