PolicyMe Cortex 🛡️

AI-Powered Insurance Intelligence Platform - Multi-agent fraud detection, risk assessment, and automated underwriting built with React + FastAPI + Gemini AI

🌟 Overview

PolicyMe Cortex is a cutting-edge InsurTech platform that leverages AI to revolutionize insurance operations. Built with React (frontend) and FastAPI + Gemini AI (backend), it automates claims processing, underwriting, fraud detection, and policy management.

Key Features

• 🔍 Claims Hawk-Eye: AI-powered claim analysis with fraud detection (0-100 risk scoring)
• 📊 Underwriting AI: Automated risk assessment and premium calculation
• ❓ Policy Q&A (RAG): Vector-based policy document search with semantic understanding
• ✅ Smart Validator: Rule-based claim validation pipeline
• 🔄 Agent Workflow: Multi-agent orchestration for complex decision-making
• 🛒 Smart Shopper: Voice-to-application quote generation

🏗️ Architecture

polcyme-cortex/
├── api/                    # FastAPI Backend
│   ├── main.py            # Main API with Gemini AI integration
│   └── requirements.txt   # Python dependencies
├── src/                   # React Frontend  
│   ├── App.js            # Main application component
│   ├── index.js          # Entry point
│   └── style.css         # Styles
├── public/               # Static assets
└── vercel.json          # Deployment configuration

🚀 Quick Start

Prerequisites

• Node.js 18+
• Python 3.9+
• Gemini API Key (get from Google AI Studio)

Local Development

1. Clone the repository

git clone https://github.com/sghosh685/policyme-cortex.git
cd policyme-cortex

2. Install frontend dependencies

npm install

3. Set up backend

cd api
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt

4. Configure environment variables

# Create .env in api/ directory
echo "GEMINI_API_KEY=your_gemini_api_key_here" > api/.env

5. Run the application

Terminal 1 (Backend):

cd api
uvicorn main:app --reload --port 8000

Terminal 2 (Frontend):

npm start

Open http://localhost:3000 🎉

🌐 Deployment

Deploy to Vercel

1. Install Vercel CLI

npm i -g vercel

2. Deploy

vercel

3. Set environment variables in Vercel Dashboard

• Go to Project Settings → Environment Variables
• Add GEMINI_API_KEY with your API key

4. Redeploy

vercel --prod

🛠️ Tech Stack

Frontend

• React 18 - UI framework
• Tailwind CSS - Utility-first styling
• Axios - HTTP client for API calls
• React Hooks - State management

Backend

• FastAPI - Modern Python web framework
• Google Gemini AI - LLM for claims analysis and NLP
• Pydantic - Data validation
• Uvicorn - ASGI server

📡 API Endpoints

Claims Analysis

POST /api/claims/analyze
Content-Type: application/json

{
  "incidentData": {
    "location": "Main St, Toronto",
    "dateTime": "2025-12-05T14:00:00Z",
    "description": "Rear-end collision",
    "injuries": false,
    "propertyDamage": true,
    "claimedAmount": 4500
  },
  "policyId": "POL-001"
}

Response:

{
  "claim_id": "CLM-20251205140000",
  "status": "approved",
  "fraud_score": {
    "score": 15.0,
    "risk_level": "Low",
    "confidence": 0.85,
    "indicators": []
  },
  "ai_analysis": {
    "validity": "valid",
    "recommendation": "auto_approve",
    "estimated_payout": 4500,
    "red_flags": [],
    "reasoning": "Standard claim with no red flags"
  }
}

Dashboard Stats

GET /api/dashboard/stats

Health Check

GET /health

🔐 Environment Variables

Variable	Description	Required
GEMINI_API_KEY	Google Gemini AI API key	Yes
REACT_APP_API_URL	Backend API URL	No (defaults to localhost)

📈 Features in Detail

Claims Hawk-Eye

• Real-time fraud risk scoring (0-100 scale)
• AI-powered claim validity assessment
• Automated approval for low-risk claims
• Multi-factor fraud indicators

Underwriting AI

• Extract structured data from broker notes
• Dynamic risk profiling
• Automated premium calculation
• Medical condition analysis

Policy Q&A (RAG)

• Upload policy PDFs
• Vector-based semantic search
• Natural language Q&A
• Citation of policy sections

Smart Validator

• Hard rule validation (policy status, limits)
• Soft rule AI checks (exclusions)
• Multi-step pipeline visualization
• Instant feedback on claim validity

Agent Workflow

• Multi-agent orchestration
• Document ingestion → Investigation → Adjudication
• Automated consensus building
• Risk-based routing

Smart Shopper

• Voice-to-text application intake
• Real-time form population
• Instant quote generation
• Conversion optimization

🎯 Roadmap

• Add PostgreSQL database integration
• Implement user authentication (JWT)
• Build admin dashboard
• Add real-time notifications
• Integrate with insurance carrier APIs
• Mobile app (React Native)
• Advanced analytics and reporting

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📝 License

MIT License - see LICENSE for details

👤 Author

Saikat Ghosh

• GitHub: @sghosh685
• Portfolio: [Your Portfolio URL]
• LinkedIn: [Your LinkedIn]

🙏 Acknowledgments

• Google Gemini AI for LLM capabilities
• Vercel for seamless deployment
• React community for amazing tools

---

⭐ Star this repo if you find it useful!

Live Demo: policyme-cortex.vercel.app