StakeFit

FitStake is rethinking how online fitness competitions work. We bring together AI verification and blockchain settlement to make contests fair, transparent, and rewarding.

Here’s how it works:

• Join a competition by staking crypto.
• You perform workouts - pushups, squats, curls, anything
• Our AI tracks 33 body landmarks in real time. It scores you on depth, tempo, and consistency.
• Once you finish, results are sent on-chain.
• Smart contracts automatically split the prize pool among the winners.

Why this matters:

• AI checks, so no bias, no human in loop.
• On-chain payouts, so results are trustless and transparent.
• Mobile-first UX makes it easy to play from anywhere, without lag.

What’s next: Wearable device support ZK-proof verification for privacy SDKs so other apps can plug into the platform

Vision: Make fitness competitions trustless, transparent, and rewarding. A place where sweat becomes stakeable, workouts become verifiable, and every rep counts.

Frontend:

• Next.js + TypeScript for a responsive, mobile-first UI.
• We integrated MediaPipe directly in-browser, tracking 33 pose landmarks at ~10 FPS.
• A custom scoring algorithm weights Range of Motion (40%), Tempo Control (35%), and Form Consistency (25%).

Backend + Contracts:

• Solidity smart contracts (deployed via Hardhat) handle staking, competition logic, and prize distribution.
• Payouts are automatic and transparent: for example, 50/30/20 splits for the top 3 performers.
• We used wagmi + viem to bridge frontend interactions with the blockchain.

Hacky bits we’re proud of:

• Instead of running a heavy ML backend, we pushed pose detection fully client-side with MediaPipe. This keeps costs near zero and makes scaling trivial.
• Built a lightweight smoothing layer to stabilize jittery landmark data, so fast reps don’t break detection.
• To simulate load, we ran competitions on Sepolia testnet, stress-testing both contract gas efficiency and the pose pipeline.

Biggest takeaway: we merged AI and Web3 together into one loop.