IntelliCredit v2

📺 Watch Demo Video

Next-Generation Credit Appraisal Engine v2, completely rebuilt to use an optimized single data pipeline with Azure Document Intelligence and Azure OpenAI GPT-4o.

Key Features of V2

1. Intelligent Ingestion: Users upload unstructured financial documents (PDFs, images, Excel). Azure Document Intelligence extracts tables, text, and layout information.
2. "Voice-to-Insight" Field Notes: Credit Officers can use the built-in browser microphone (Web Speech API) to dictate immediate field observations and risk constraints before analysis.
3. Deep AI Underwriting: Azure OpenAI (GPT-4o) processes the complete document context + officer notes. It returns a fully structured JSON response containing:
   • Score & Grade: 0-100 total score securely rounded.
   • 5 C's Analysis: Character, Capacity, Capital, Collateral, and Conditions.
   • Document Consistency Audit: Cross-verifies PAN, GST turnover vs P&L revenue, and Director names across multiple documents. Highlights critical mismatches.
   • Lending Recommendation: Final go/no-go, suggested limits, and interest rates.
4. Dynamic Recalculation: Officers can iteratively add new insights and command the AI to re-evaluate the profile, instantly syncing the updated scores to the dashboard.

System Architecture

The V2 architecture operates on a modern, decoupled pipeline:

• Frontend: Next.js 14, React Hook Form, TailwindCSS, lucide-react icons. Completely SSR/CSR optimized.
• Backend: Python FastAPI, asynchronous endpoints, SQLAlchemy 2.0.
• Database: PostgreSQL for structured data persistence (applications, documents, analyses).
• AI Integrations:
  • @azure/ai-document-intelligence for robust OCR.
  • openai python SDK pointing to Azure GPT-4o deployments.

1. Prerequisites

• Docker and Docker Compose
• Node.js (v18+) for local frontend development
• Python 3.10+ for local backend development
• Azure OpenAI and Azure Document Intelligence API keys

2. Environment Configuration

Copy the .env.example file to .env in the root directory:

cp .env.example .env

Populate the keys:

AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
AZURE_OPENAI_API_KEY=your_key
AZURE_OPENAI_DEPLOYMENT_NAME=gpt-4o
AZURE_DOCUMENT_INTELLIGENCE_ENDPOINT=https://your-resource.cognitiveservices.azure.com/
AZURE_DOCUMENT_INTELLIGENCE_KEY=your_key
DEMO_MODE=false

Note: Set DEMO_MODE=true to skip actual Azure API calls and use the injected Pujari Industries test data.

3. Run with Docker Compose

From the root directory, simply run:

docker-compose up --build

This will start:

• PostgreSQL Database on port 5432
• FastAPI Backend on port 8000 (http://localhost:8000/docs)
• Next.js Frontend on port 3000 (http://localhost:3000)

4. Seeding Demo Data (Optional)

To instantly populate a full "Pujari Industries" application for testing the dashboard:

# Enter the backend container
docker-compose exec backend bash

# Run the seeder
python seed.py

Refresh the frontend dashboard at http://localhost:3000 to view the comprehensive credit report.