Cross-Chain Prediction Market with AI Powered Odds

Demo

About

The problem it solves ## Problem It Solves Traditional betting platforms are centralized and opaque, allowing companies to set unfair odds, delay or refuse payouts, and block users without transparency. Users must trust a closed system with little visibility into how odds or results are handled. --- ### What This Project Enables: **Decentralized, Trustless Betting With Chainlink** All bets and payouts are handled by smart contracts, ensuring fairness and eliminating centralized control. **AI-Driven, Data-Backed Odds Powered by Chainlink Functions** Odds are calculated by a trained AI model using real data and updated on-chain via Chainlink Functions—ensuring accuracy and eliminating bias. **Tamper-Proof and Verified Data** Odds are set using a trained AI model combined with real-time data fetched from Polymarket by an autonomous AI agent. Chainlink Functions brings these insights on-chain, ensuring accuracy, transparency, and zero manual bias. **Fully Transparent and Auditable Payouts** All transactions are on-chain and auditable, giving users complete trust in the system.

How it's made

Challenges we ran into ### Challenges & How I Overcame Them ## 1. **No Access to Testnet Tokens (Especially Avalanche Fuji)** I quickly ran into a problem when I didn’t have testnet tokens to deploy or test my contracts — particularly **AVAX on Fuji** and **Sepolia ETH** because I didnt plan to purchase LINK. **Solution:** - For Sepolia ETH, I found faucets through online communities. - But for Fuji AVAX, it was much harder — I even ran into scammers while asking around in the Avalanche Discord group. - Eventually, I decided to **buy 1 LINK** and used **Chainlink's official website** to get the testnet assets I needed. - This helped me move forward with building and testing cross-chain logic. --- ## 2. **Struggled with Chainlink Functions** I had never used Chainlink Functions before, and initially I couldn't even get it to pass data from off-chain to my smart contract, even the AI cannot help as it gave the wrong way, I assumed they are not trained well in this part. **Solution:** - I followed the official **Chainlink Functions documentation** and successfully deployed the example first. - When my co didn’t work, I debugged the issue by: 1. Realizing the **ngrok tunnel** I used for local dev caused dynamic URLs, which broke the data fetch. I fixed this by using a static domain. 2. Looking at recent fulfillment status and logs, I noticed failures in computation or callbacks. This helped me pinpoint whether the issue was with data fetching or smart contract execution. --- ## 3. **Learning & Debugging Chainlink CCIP** CCIP was the most complex part — it’s not just contract logic, it’s **cross-chain messaging with tokens and data**. The AI still didnt work here. **Solution:** - I started by completing the **official CCIP "send tokens with data" example** to learn the basics. - Then I customized it for my project: my messages included a `struct`, and I needed to forward that data to another contract on the receiving chain. - Despite following the guide, things still didn’t work, and debugging cross-chain issues was tough. - Eventually I learned to use tools like **CCIP Explorer** and **Tenderly** to trace message delivery and execution paths. - I hit a wall when nothing seemed wrong — but then got help from the **Chainlink Discord community**, who found that the **gas limit was too low on the receiving chain**. - That insight gave me confidence that my logic was correct, and I was finally able to complete the cross-chain flow. --- Each of these hurdles was tough, but working through them helped me build deeper understanding and confidence in Web3 development — especially with Chainlink’s powerful tools like Functions and CCIP.