VeilScore: Private credit scoring with zero-knowledge machine learning. Generate verifiable credit scores on-chain without revealing sensitive financial data. Built on Mantle with Groth16 proofs.
VeilScore revolutionizes credit scoring by combining zero-knowledge machine learning (ZKML) with blockchain technology. Traditional credit scoring requires users to expose sensitive financial data, creating privacy risks and data breaches. VeilScore solves this by enabling private, verifiable credit scoring on-chain without revealing any personal information.
The Problem:
Current credit scoring systems force users to share sensitive financial data with lenders, creating privacy vulnerabilities and potential for data misuse. There's no way to prove creditworthiness without exposing personal information.
The Solution:
VeilScore uses zero-knowledge proofs (Groth16) to generate verifiable credit scores on blockchain. Users input their financial data, which is processed by a high-performance LightGBM machine learning model (trained on real LendingClub data with 99%+ accuracy). The system generates a zero-knowledge proof that verifies the credit score calculation without revealing the underlying data.
Key Features:
- Privacy-First: Financial data never leaves the user's control or is exposed on-chain
- Verifiable: Groth16 proofs ensure score authenticity without data disclosure
- High Performance: LightGBM model achieves >0.90 AUC on real-world data
- Explainable: SHAP-based analysis shows top 3 feature impacts
- On-Chain Verification: Smart contracts on Mantle Network verify proofs
- Production Ready: Full-stack application with FastAPI backend, React frontend, and comprehensive test coverage
Technical Stack:
- Backend: Python, FastAPI, LightGBM, EZKL for ZK proofs
- Frontend: React, TypeScript, Wagmi, Viem
- Blockchain: Solidity smart contracts on Mantle Network
- Infrastructure: Docker, CI/CD with GitHub Actions
Live Deployment:
- Web App: https://veilscore.onrender.com/
- Smart Contracts: Deployed on Mantle Sepolia Testnet
- Fully functional and ready for use
VeilScore demonstrates the future of private DeFi applications, where users can prove creditworthiness without sacrificing privacy.
**Week 1: Core Development** **Phase 1: ML & ZK Foundation (Days 1-2)** - Processed 200K LendingClub samples, trained LightGBM model (99%+ accuracy) - Integrated EZKL pipeline for Groth16 zero-knowledge proofs - Generated Solidity verifier contract for on-chain verification **Phase 2: Backend & Smart Contracts (Days 3-4)** - Built FastAPI REST API with /prove endpoint - Developed PrivateCreditLending.sol smart contract - Deployed contracts to Mantle Sepolia Testnet - Achieved >95% test coverage **Phase 3: Frontend & Integration (Days 5-6)** - Created React frontend with Web3 integration (Wagmi/Viem) - Built multi-step wizard for credit scoring workflow - Connected all components (frontend → API → blockchain) - Containerized with Docker **Phase 4: Deployment & Testing (Day 7)** - Deployed live application to Render - Set up CI/CD pipeline with GitHub Actions - Created comprehensive E2E test suite - Final bug fixes and documentation **Key Achievements:** ✅ Full-stack application (React + FastAPI + Solidity) ✅ Production-ready ML model with ZK proofs ✅ On-chain verification on Mantle Network ✅ Live deployment with CI/CD ✅ Comprehensive testing (>95% coverage)
Pre-seed stage. Self-funded development during hackathon. Exploring seed funding opportunities to accelerate development, expand team, and scale infrastructure for production deployment.