AI-Driven Cross-Selling Insurance Predictor | UAB Academic Project
Prompt + Response Writing (SFT)Built and tested an end-to-end machine learning workflow using advanced prompt engineering to generate, review, and debug Python code. Focused on improving model performance under severe class imbalance by adjusting pipeline logic and evaluation thresholds. • Used Cursor, Gemini, and Ollama to prompt for code generation and iterative debugging. • Implemented a cost-sensitive Logistic Regression approach for 1:5 class imbalance. • Tuned probability thresholds to dramatically improve recall from near-zero to >99%. • Validated results by running and testing the full Python pipeline end-to-end.
2025 - 2025