Real time API’S Credit Score for Loan underwriting.
For a non-banking finance company
CLIENT & PROBLEM STATEMENT
- Non-banking finance company offering SME loans needs ML-driven loan underwriting scores.
- Requirement for real-time scoring via UI and APIs, with batch training of ML models.
APPROACH
- Established secure Python-based ML APIs for efficient data transfer from client systems.
- Developed ML Scorecards, a custom framework for creating and managing underwriting models.
- Integrated historical loan data, SME financials, and market data to train robust logistic regression and tree-based models.
- Implemented feature importance analysis to ensure model transparency and regulatory compliance.
- The incremental learning approach allowed the model to continuously improve based on incoming data, ensuring accuracy and relevance in real-time scoring.
- Deployed models in a cloud environment for batch retraining, while exposing scoring endpoints for real-time API and UI integration.
SOLUTION & OUTPUT
- ML Scorecards using APIs provide accurate, real-time loan underwriting scores via UI and APIs.
- Faster, data-driven SME loan approvals with reduced risk, enhancing portfolio quality.
- The solution streamlined real-time data transfer and scoring through API integration, allowing continuous batch retraining and real-time interaction via UI.
- This resulted in faster decision-making and improved underwriting accuracy, meeting both operational and regulatory needs.
