Teleo

Inspiration

The gig economy is broken by a lack of trust. Freelancers fear "ghosting" and non-payment after doing the work, while Clients fear paying for subpar results. Traditional platforms try to fix this with human mediation, but that leads to high fees (20%+), slow disputes (weeks), and bias.

We asked: What if we replaced the middleman with Code and AI?

Teleo (derived from the Greek Telos, meaning "End Goal") focuses purely on the output. By combining MNEE Stablecoin for programmable money with AI Agents for objective verification, we created a protocol where trust is automated, not negotiated.

Teleo is the first of its kind AI Arbitrated Work Platform

What it does

Teleo is a decentralized marketplace where payouts are guaranteed by code, not promises.

Smart Escrow: Clients post a job and lock MNEE tokens into a Smart Contract. The funds are physically secured on-chain.
AI Arbitration: Instead of a human manager, an AI Agent reviews the submission. It analyzes code, text, or data against the client's requirements in seconds.
Instant Settlement:
Pass: The contract automatically releases the MNEE to the freelancer.
Fail: The freelancer gets specific AI feedback to iterate and resubmit.

We built a unique "God Mode" simulation for this hackathon, allowing judges to switch between Client and Freelancer personas instantly to experience the full lifecycle without needing multiple wallets.

How we built it

We built Teleo as a bridge between Web3 infrastructure and Web2 usability:

Blockchain: We wrote a custom Solidity Escrow contract deployed on Sepolia (and configured for Mainnet). It interacts directly with the MNEE Stablecoin contract for deposits and transfers.
Frontend: Built with Next.js 14 and Ethers.js. We focused heavily on Mobile Responsiveness and created a "Runtime Config" system to toggle seamlessy between Sepolia (Demo) and Mainnet (Production).
The AI Judge: A Python FastAPI backend acts as the oracle. It accepts files/text, uses *gemini to evaluate the deliverables against the job description, and signs a transaction with its own private key to trigger the on-chain release.
Data: We used MongoDB to index blockchain events, enabling features like "Submission History" and "Tag Filtering" that raw blockchain data can't provide easily.

Challenges we ran into

Event Log Parsing: We initially struggled to sync the "Job ID" between our database and the Blockchain. The script was grabbing the Amount (50 ETH) instead of the ID from the event logs because of how un-indexed parameters work. We fixed this by implementing strict ABI parsing using contract.interface.parseLog.
Environment Variables in Docker: Deploying to Azure broke our contract connections because Next.js bakes variables in at build time. We engineered a custom entrypoint.sh script to inject Azure App Settings into the browser at runtime (window.env), keeping our secrets safe while allowing dynamic configuration.
Gas Management: Our AI Judge kept running out of ETH for gas fees! We implemented a robust error handling system that alerts us if the Judge's wallet is empty before attempting a verdict.

Accomplishments that we're proud of

True "God Mode": We created a frictionless demo experience where judges can simulate a multi-user economy using a single wallet address, solving the "Empty Wallet" problem for testing.
Mainnet Ready: While we demo on Sepolia, our architecture includes a "Network Toggle" that instantly switches the entire app to the Ethereum Mainnet MNEE contract (0x8cc...), proving technical compliance.
Interactive Feedback Loop: We didn't just build a "Pass/Fail" switch. We built a full submission history where freelancers can see exactly why the AI rejected them and improve their work.

What we learned

Programmable Money is Powerful: working with MNEE showed us how stablecoins are superior to fiat for escrow—there are no banking delays, just code execution.
The Complexity of "Trustless": Designing a system where neither the Client nor the Admin can steal the funds required rigorous logic in our Solidity modifiers (onlyJudge, refundClient).
Hybrid Architecture: We learned that the best DApps aren't 100% on-chain. Combining the speed of a Web2 DB (MongoDB) with the security of Web3 (Ethereum) creates the best user experience.

What's next for Teleo

From God Mode to Real Auth: We plan to move from our simulation login to full Web3 Authentication (SIWE), allowing any freelancer to create a profile and generate shareable "Contract Links" to invite clients.
Multimedia & Async Judging: Currently, we handle text and code. We will expand the AI Judge to process images and videos asynchronously, enabling design and content creation jobs with longer verification times.
Complex Deliverables: We want to support "Milestone-Based" payouts. This includes time-based partial releases and specific checklist-based verification to cater to complex, long-term client projects.