Inferra

Our project is an agentic AI-powered prediction market system inspired by Polymarket. It enables autonomous agents to analyze real-world data, form opinions on prediction events (e.g., “Will BTC hit 200k by 2025?”), and execute trades on-chain without human intervention.

The system involves two key components:

Research Agent – uses AI models to browse web data, gather insights, and reason about probabilities.

Execution Agent – interacts with the blockchain, submitting buy/sell orders based on the research output.

Together, they create a fully autonomous decision loop — data → reasoning → action — blending AI autonomy with DeFi prediction markets.

This opens a new frontier called Agentic DeFi, where intelligent systems participate in markets, generate alpha, and adapt over time through feedback loops.

The project is built with a combination of AI reasoning frameworks, smart contracts, and web3 infrastructure.

Frontend: React + Tailwind CSS for a clean, modular UI displaying markets, agent actions, and results.

Backend / AI Layer: Node.js + LangChain/OpenAI API (for autonomous reasoning and data synthesis).

Agent Framework: AutoGPT or custom-built multi-agent architecture for the Research and Execution agents.

Blockchain Integration: Smart contracts on Polygon or Base, using Solidity and Ethers.js for market interaction.

Off-chain Data: AI retrieves live market sentiment, crypto price feeds, and news data via APIs (like CoinGecko, Google Search, etc.).

Execution: Research Agent passes reasoning output via a message bus (like Redis or local event queue) to the Execution Agent, which performs on-chain trades.

We also used Push Protocol for agent communication alerts and Lighthouse/IPFS for decentralized storage of agent logs.

One notable hack: agents can cross-check multiple AI models to ensure reasoning accuracy before trading, simulating human-like “research verification.”