Client Overview
Client: Leading Non-Banking Financial Company (NBFC) in India
Sector: Retail and MSME Lending
Challenge: Reduce NPAs and automate credit risk decisions
Duration: 4 months (end-to-end delivery)
Business Problem
The client relied on rule-based underwriting, resulting in high delinquency and suboptimal credit decisions. The NBFC needed a scalable ML platform leveraging historical and bureau data to improve credit risk prediction and reduce NPAs.
Our Solution
- Data Integration: Integrated 5 years of internal lending data and 3 credit bureau APIs
- Feature Engineering: Created 150+ features across bureau, behavioral, and derived segments using PySpark
- Modeling: Developed and evaluated XGBoost, Random Forest, and Logistic Regression models
- Deployment: Real-time API scoring integrated with LOS and field officer dashboards
- Monitoring: Power BI dashboards for drift, delinquency, and model insights
Impact Delivered
Impact Area | Outcome |
---|---|
Approval Accuracy | 26% improvement in bad loan detection |
Turnaround Time | Reduced from 36 hrs to < 6 hrs |
NPA Reduction | Projected 18–24% drop in new NPAs |
Scalability | Supports over 1 million annual loan applications |
Compliance | SHAP explainability met audit & regulator standards |
Technology Stack
Data Engineering: PySpark, Airflow
Modeling: Scikit-learn, XGBoost, SHAP
APIs: Flask, FastAPI
Dashboards: Power BI
Infrastructure: AWS EC2, S3, Lambda
Client Testimonial
“The ML-based credit risk platform delivered by Durga Analytics has transformed our underwriting process. The predictive accuracy and real-time decisioning capability have not only improved our portfolio quality but also empowered our field officers to make faster and smarter lending decisions.”
— Chief Risk Officer, NBFC Client
Next Steps
We’re now extending the platform to include top-up loan prediction, prepayment risk, and UPI-based behavioral scoring.
Want to revolutionize your credit decisioning with AI & ML? Talk to Our Experts