At Tata AIG, I built machine learning pipelines for health insurance fraud detection, including a heterogeneous GraphSAGE model that identified ₹14 lakh in additional fraudulent claims. The work also surfaced provider and producer blacklists with 50% overlap against existing SIU lists.
I designed health fraud models that achieved ROC above 85 and automated work equivalent to 500 hours per month. I also deployed an ETF selection and allocation strategy that achieved 4% excess XIRR over the actual portfolio in backtesting.
On RoamMate, I developed a travel-planning agent using LangGraph and FastAPI, with destination rankings based on YouTube engagement data and Wikidata. I also worked as a Data Science Intern at OYO, automating reporting and SEO analytics.
