The problem OpenBar solves

Freelance platforms like Upwork and Fiverr control a $1.5T market but charge 10–20% in fees, hide how disputes are resolved, and keep all funds and decisions inside a single company. Buyers and freelancers must trust that platform decisions are fair, with no way to audit them. Escrow and dispute logic are opaque and centralized: disputes are decided behind closed doors, with no public record or explanation. Funds sit in platform accounts, so users depend on a single point of trust and can’t move to alternatives without losing history. OpenBar addresses this with transparent, on-chain freelance escrow: low fees (~1%), peer validation with votes logged on Hedera Consensus Service, and payouts handled via Hedera’s native Transfer and Schedule services. Users get verifiable fairness and auditable outcomes instead of trusting a central authority.

Challenges we ran into

Wallet-signed transactions. Getting the frontend to sign Hedera transactions was the hardest part. The backend builds transactions with the Java SDK, but the frontend uses WalletConnect and the JS SDK. When passing frozen transaction bytes between SDKs, signTransaction + Transaction.fromBytes() caused INVALID_SIGNATURE because protobuf bytes changed on the way. It took several hours of debugging with StackOverflow and ChatGPT to understand this. The fix was to use signAndExecuteTransaction so the wallet receives raw base64 bytes and submits directly to Hedera, avoiding cross-SDK serialization.
Hedera Consensus Service (HCS). Making HCS work for the audit trail was tricky. Setting up topics, message format, and error handling when publishes fail required a lot of trial and error. We also had to coordinate lifecycle events (TASK_CREATED, VOTE_CAST, DISPUTE_RESOLVED) so the audit trail stays consistent.
Schedule Service for auto-refunds. Wiring Hedera’s Schedule Service for time-based refunds if no freelancer claims in time took careful handling of expiry times, admin keys, and deleting schedules when a job is claimed.
Coordinating three Hedera services. Using Transfer, Schedule, and Consensus together in the right order and error paths, while keeping wallet-signed vs operator-signed transactions clear, added complexity to the flow.
Peer validation flow. Designing the dispute flow so validators (with reputation ≥ 50) vote correctly, reaching quorum, and paying the right party on time required careful state and edge-case handling.

Use of AI tools and agents

We used AI tools to build and improve the project:

• UI and layout, Generating and refining React/Tailwind components for the landing page, job views, and modals

• Debugging, Tracking down errors such as cross-SDK transaction signing (Java → JS/WalletConnect), with guidance from AI to locate fixes

• Documentation, Adding Javadoc for backend code, JSDoc for frontend, and improving the README for clarity

• Project structure, Refactoring, extracting shared utilities, and removing dead code with AI help

• Hackathon prep, Using Hedera hackathon skills to evaluate the submission, refine the pitch, and improve the README