An autonomous, TEE-secured DeFi agent built on 0G Compute to mathematically eliminate MEV front-running and 'Black Box' AI risks.
AutoYield AI is a next-generation, verifiable finance protocol built specifically for the 0G APAC Hackathon (Track 2: Agentic Trading Arena). We are bridging the gap between autonomous AI trading and on-chain security by eliminating the risks of MEV front-running and opaque AI execution.
The Problem: Currently, AI trading agents operating in DeFi suffer from two major flaws: their off-chain decision-making is a "black box" that users must blindly trust, and their on-chain execution is vulnerable to MEV bots that front-run profitable trades. Furthermore, naive vault rebalancing often causes complete transaction failure during protocol liquidity crunches.
The Solution: AutoYield AI utilizes a fully autonomous, TEE-secured DeFi agent that verifies risk-adjusted yield strategies. By leveraging 0G's modular infrastructure, we ensure that the AI's "thought process" is verifiable, while the actual trade execution remains completely hidden from front-runners.
Architecture & 0G Integration:
0G Compute (TEE Execution): The core of our innovation. When our decision engine calculates the optimal APY, it executes inside a 0G Compute Trusted Execution Environment. This seals the market data and generates a cryptographic execution proof, guaranteeing the AI wasn't tampered with by developers.
Decentralized Oracle Ingress: To prevent data spoofing, the AI agent bypasses centralized Web2 APIs and natively digests on-chain data directly from decentralized oracles (Chainlink/Pyth).
Upgradeable Smart Contracts (0G Chain): We utilize modern ERC-4626 standard yield-bearing vaults (AutoYieldVaultUpgradeable.sol) managed by a central strategy router (StrategyManager.sol). We engineered "Delta Rebalancing" to survive real-world liquidity traps.
Backend Keeper Infrastructure: A robust Node.js transaction queue powered by PM2 prevents nonce-collisions and perfectly handles user load.
During the 0G APAC Hackathon timeline, our team evolved this project from an initial concept into a full-stack, mainnet-ready Web3 application:
Smart Contracts: Wrote and deployed an ERC-4626 Vault architecture to the 0G Newton Testnet. Transitioned from rigid monoliths to OpenZeppelin Upgradeable Proxy contracts to ensure future-proofing.
AI & Math: Engineered a basis-point (BPS) mathematically accurate decision engine that balances live APY yields against calculated protocol risk scores.
Enterprise DevOps: Overhauled the backend to include persistent transaction queuing, preventing ethers.js nonce-collisions under load, and added PM2 containerization for 24/7 uptime.
Security Hardening: Ripped out centralized Web2 data feeds and built a decentralized oracle service to ensure the AI's input is as secure as its 0G Compute TEE output.
Bootstrapped. We are actively seeking ecosystem grants and seed funding to take AutoYield AI to 0G Mainnet and expand our verifiable trading strategies.