JP Morgan Credit Default Predictor
This project was developed as a spin-off of the JP Morgan Chase & Co. Quantitative Research Virtual Internship.
The goal was to build a reliable model that predicts the probability of a customer defaulting on a credit payment — a key task in minimizing financial risk and improving lending decisions.
The work combines two parts:
- Analysis & Model Selection: a full Jupyter Notebook workflow, including EDA, feature engineering, handling class imbalance, and hyperparameter tuning. I compared a classic XGBoost model with a Deep Learning approach (TensorFlow/Keras) to identify the best performer.
- Deployment as a Web App: the chosen model was integrated into a production-style FastAPI application, containerized with Docker, and deployed to Render for real-time credit risk prediction.