AURA (Automated Underwriting & Risk Assessment) is an intelligent decision-support platform designed to modernize and streamline insurance underwriting. It combines deterministic business rules with probabilistic machine learning to deliver transparent, fair, and regulation-compliant decisions.
AURA uses a Hybrid Decision Architecture, blending:
- Deterministic Business Rules -- for strict regulatory compliance and sanity checks\
- Probabilistic Machine Learning -- for identifying non-linear risk patterns such as DTI ratios and lifestyle correlations
This ensures explainability without sacrificing predictive performance.
AURA is built around interpretability, auditability, and modular design. The system evaluates insurance applications and returns:
- ACCEPT
- REJECT
- MANUAL_REVIEW
Each decision is paired with explainability metadata so underwriters understand why the decision was made.
A weighted decision model blending: - 40% Business Rules - 60% AI Model Output
A responsive SPA user interface.
FastAPI-based SOA.
Automatic knockout rules.
Transparent reasoning metadata.
AURA/
├── data/
├── frontend/
│ └── index.html
├── models/
│ └── saved/
│ └── risk_model.pkl
├── src/
│ └── autosure/
│ ├── domain/
│ ├── interface/
│ ├── rules/
│ └── services/
├── tests/
├── datagen.py
├── train_model.py
├── run.py
└── requirements.txt
pip install -r requirements.txt
python datagen.py
python train_model.py
python run.py
Open frontend/index.html.
Compliance rules and knockout logic.
Random Forest classifier.
Final Score = (Rule Score * 0.4) + (AI Score * 0.6)
Score Decision Action
0-40 ACCEPT Auto-approve 41-75 MANUAL_REVIEW Human review 76-100 REJECT Decline
- /docs (Swagger UI)
- /redoc (Redoc)
pytest tests/
Built for the 2025 Insurance Tech Challenge.