VIVIRITY Intelli-Credit
A comprehensive credit risk intelligence system leveraging machine learning to assess borrower risk, predict defaults, and provide actionable insights for financial institutions. Features an interactive dashboard, real-time risk scoring, and explainable AI for regulatory compliance.
Problem
Traditional credit scoring models fail to capture complex patterns in borrower behavior, leading to high default rates and missed opportunities for creditworthy individuals.
Solution
Built an ensemble ML system combining gradient boosting, random forests, and neural networks with SHAP explainability to deliver accurate, transparent credit risk assessments.
Architecture
- Data pipeline: ETL from multiple credit bureau sources
- Feature engineering: 200+ derived features from transaction history
- Ensemble model: XGBoost + Random Forest + Neural Network
- API layer: FastAPI with async endpoints
- Dashboard: Interactive risk visualization with D3.js
- Explainability: SHAP values for regulatory compliance
Results
92% accuracy in default prediction, 40% reduction in false positives, secured 2nd place at IIT Hyderabad YUVAAN 2026.
FAQs
What dataset was used?
The model was trained on a combination of synthetic credit data and publicly available lending club data augmented with custom feature engineering.
Can this be deployed in production?
Yes, the system is containerized with Docker and includes a FastAPI backend suitable for production deployment with proper infrastructure.