End-to-end AI/ML projects, from data preprocessing to deployment
Machine learning–driven credit risk analytics platform integrating loan-level scoring with portfolio monitoring, stress testing, and ECL simulation. Built with Angular, Flask, and LightGBM.
This project demonstrates a structured credit risk analytics framework that combines probability of default (PD) estimation with portfolio-level risk intelligence. The system integrates model inference, explainability, stress testing, migration analysis, vintage monitoring, and expected credit loss (ECL) estimation within a governed architecture.
Rather than focusing solely on individual loan predictions, the platform illustrates how machine learning outputs can be aggregated, monitored, and stress-tested to support capital impact analysis and risk governance in regulated financial environments.
Machine learning–based transaction monitoring and fraud risk analytics platform. Built with Streamlit and LightGBM.
This project demonstrates a structured fraud detection framework built using Streamlit and LightGBM. The application integrates model inference, monitoring dashboards, and explainability tools into a unified analytical interface.
It simulates batch scoring, real-time transaction monitoring, alert threshold optimization, and SHAP-based model explainability. The platform is designed to illustrate how machine learning can support operational fraud monitoring, alert governance, and system-level stability analysis.