Prism finance
AI-powered DeFi lending protocol that analyzes on-chain behavior using TensorFlow to provide personalized interest rates. Good borrowers save up to 25% on rates while lenders earn risk-adjusted yields
Videos
Description
Pitch Deck: https://gamma.app/docs/Prism-Finance-h1qs61sp0ew10ux
Prism Finance revolutionizes DeFi lending by bringing TradFi credit scoring to blockchain. Unlike Aave or Compound where everyone pays the same rate, Prism Finance uses AI to analyze on-chain behavior and rewards good borrowers with up to 25% lower interest rates.
Core Innovation:
- 🤖 AI Credit Scoring: TensorFlow neural network analyzes 9 on-chain metrics to calculate credit scores (0-1000) with 98% accuracy
- 📊 3 Risk-Based Pools: Choose your risk tolerance (Low/Medium/High) with different collateral ratios (150%/175%/200%)
- 💎 Personalized Rates: Better credit score = lower interest rates. Users with excellent credit (850+) save up to 25% compared to poor credit (300)
- ⚡ Built on Mantle: Modular L2 with low fees and high throughput
Example Interest Rates (60% pool utilization, Medium risk tier):
- 850 (Excellent): 6.9% APR
- 600 (Average): 9.3% APR
- 300 (New/Risky): 12.6% APR
Technical Stack:
- Smart Contracts: Solidity 0.8.20, Foundry, OpenZeppelin (121/121 tests passing)
- AI Model: Python 3.9 + TensorFlow 2.15 (98% accuracy, R² = 0.982)
- Frontend: Next.js 15 + React 19, Privy wallet integration
- Infrastructure: Mantle Network (400k+ TPS), Chainlink oracles
Live Deployments:
- Frontend: https://prism-finance-ten.vercel.app/
- AI API: https://prism-finance-ai.onrender.com/
GitHub: https://github.com/7maylord/prism-financeProgress During Hackathon
Over the course of the hackathon, the project progressed from core protocol design to full deployment. I laid the foundation by designing the smart contract architecture, implementing a multi-tier lending pool, building the credit oracle infrastructure, and developing a rate calculator with credit-based adjustments. Then i focused on AI integration, where we trained a TensorFlow neural network on 20,000 samples, achieving 98% credit-scoring accuracy, and integrated the AI API with the smart contracts alongside an automated credit score updater bot. Thereafter, I delivered a production-ready Next.js frontend with five pages, integrated Privy wallet authentication, added real-time health factor monitoring, and completed a comprehensive test suite spanning 121 tests. Finally, I deployed the smart contracts to Mantle Sepolia, launched the frontend on Vercel, deployed the AI API on Render, set up four automation bots, and finalized full project documentation.
Tech Stack
Fundraising Status
Seed