UD — Credit Underwriting Engine — Technical Spec

In Review

SakshamAI

Saksham is ready

Ask me to help draft, review, or generate content

Reflections

# Understanding Document — Credit Underwriting Engine

## 1. Overview
This UD captures the technical architecture and functional specification of the automated credit underwriting engine.

## 2. Functional Specification
The credit underwriting engine evaluates loan applications using a combination of bureau data, internal scoring models, and policy rules.

### 2.1 Scoring Model
- Primary: CIBIL score (weight: 40%)
- Secondary: Internal behavioral score (weight: 30%)
- Tertiary: Income-to-EMI ratio (weight: 30%)

## 3. Technical Specification
### 3.1 Technology Stack
- Rule Engine: Drools 8.x
- ML Model: XGBoost (Python, served via FastAPI)
- Database: PostgreSQL 15 for decision audit trail

## 4. Data Model
### 4.1 Input Schema
```json
{
  "applicantId": "string",
  "panNumber": "string",
  "cibilScore": "number",
  "monthlyIncome": "number",
  "existingEMI": "number",
  "loanAmount": "number",
  "tenure": "number"
}
```

## 5. Integration Points
- CIBIL API: POST /v3/credit-report
- Internal Scoring API: POST /scoring/v2/evaluate
- Drools Rule Engine: REST API

## 6. Edge Cases
- Applicant with no credit history (thin file): Route to manual underwriting
- NRI applicants: Use overseas credit report if available