JIA Finance - Quantitative Analyst
- Automated Mortgage Guidelines Interpretation leveraging OpenAI GPT 3.5, GPT 4 APIs, LangChain, FAISS/Chroma DB vectorstore for rapid ingestion & validation of documents resulting in enhanced speed & accuracy of mortgage processing.
- Engineered specialized task-specific Retrieval Augmented Generation (RAG) Q&A chains that refined query processing using a recursive tree approach combined with LLM-based memoization which significantly increased response accuracy for queries boosting zero shot accuracy by 30 %.
- Exploratory Data Analysis on over 100M Fannie Mae mortgage records using Tableau along with Data Cleaning & Feature Engineering with AWS Sagemaker & Redshift Connector, applying various other data manipulation & visualization tools for actionable insights.
- Modeled loan survival & default rates curves employing the Cox Proportional Hazard Model & Kaplan Meier Estimator & assessed the impact of various macroeconomic factors such as changes in unemployment rates & income on default & delinquency rates.
- Crafted a comprehensive cash flow model for Mortgage Backed Securities (MBS) integrating credit default, prepayment, & loss severity models with a yield curve calibrated Cox Ingersoll Ross (CIR) Interest Rate model.
