Federated Learning for Privacy-Preserving Financial Data Generation with RAG Integration

This project implements a federated learning framework combined with a Retrieval-Augmented Generation (RAG) system to generate privacy-preserving synthetic financial data.

Features

Federated Learning using TensorFlow
Privacy-preserving data generation using VAE/GAN
RAG integration for enhanced data quality
Secure Multi-Party Computation (SMPC)
Differential Privacy implementation
Kubernetes-based deployment
Comprehensive monitoring and logging
NEW: Interactive Web Demo - Try it out without setup!

Quick Demo (No Installation Required)

🚀 Live Demo: Hugging Face Spaces

The web demo allows you to:

Enter customer features and get credit score predictions
See how federated learning works
Understand privacy-preserving ML concepts

Installation

# Create virtual environment
python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

Federated Credit Scoring Demo (with Web App)

This project includes a demo where multiple banks (clients) collaboratively train a credit scoring model using federated learning. A Streamlit web app allows you to enter customer features and get a credit score prediction from the federated model.

Quick Start

Install dependencies

pip install -r requirements.txt

Start the Federated Server

python -m src.main --mode server --config config/server_config.yaml

Start at least two Clients (in separate terminals)

python -m src.main --mode client --config config/client_config.yaml

Run the Web App

streamlit run webapp/streamlit_app.py

Use the Web App

Enter 32 features (dummy values are fine for demo)
Click "Predict Credit Score" to get a prediction from the federated model
View training progress in the app
Toggle between Demo Mode (no server required) and Real Mode (connects to server)

For best results, keep the server and at least two clients running in parallel.

Project Structure

FinFedRAG-Financial-Federated-RAG/
├── src/
│   ├── api/           # REST API for server and client communication
│   ├── client/        # Federated learning client implementation
│   ├── server/        # Federated learning server and coordinator
│   ├── rag/           # Retrieval-Augmented Generation components
│   ├── models/        # VAE/GAN models for data generation
│   └── utils/         # Privacy, metrics, and utility functions
├── webapp/            # Streamlit web application
├── config/            # Configuration files
├── tests/             # Unit and integration tests
├── docker/            # Docker configurations
├── kubernetes/        # Kubernetes deployment files
└── app.py             # Root app.py for Hugging Face Spaces deployment

License

MIT

Contributing

Please read our contributing guidelines before submitting pull requests.

Demo URL: https://huggingface.co/spaces/ArchCoder/federated-credit-scoring