Finish AI Voice Platform (ongoing work)
Budget: $50 – $0 AUD
My AI-driven voice interview app is about 70% complete. The codebase—React 18.3, TypeScript, Supabase, Tailwind CSS—already handles authentication, basic interview flow, and partial OpenAI Realtime streaming. What's blocking launch is the real-time layer: WebRTC audio occasionally drops, and the transcription stream desynchronises, which then stalls the AI response pipeline.
The platform conducts voice interviews using OpenAI's Realtime API, capturing and analysing responses in real-time. Four Supabase Edge Functions handle token generation, interview processing, and notifications—these may need debugging alongside the WebRTC fixes. Audio streams at PCM16/24kHz. The React app connects to OpenAI through a Supabase Edge Function relay for security. Session management and ephemeral tokens are already implemented but may need refinement during WebRTC debugging.
The first milestone is, therefore, a rock-solid WebRTC fix. I need those dropped-connection edge cases handled, with graceful reconnection and accurate transcription timestamps, all verified in the live production environment we'll ultimately ship on Vercel.
Once the stream is stable, the scope widens:
- Finish the AI response pipeline so answers come back in real time
- Integrate Stripe for subscription billing (monthly & annual plans already defined)
- Build an admin dashboard for the interview template and user management
- Add five missing pages—results, settings, library, plus two small utility screens—styled to match existing Tailwind components
- Implement global error handling, sensible API rate-limiting, and security hardening
- Optimise bundle size and lighthouse performance before final deploy
I will provide full repository access, existing Supabase tables with 13 configured tables and RLS policies, and the existing UI/component structure using shadcn/ui. Please keep the current structure; the goal is completion, not a rewrite. Clear inline docs and a short handover guide are part of the final deliverable so my team can maintain it.
NDA required before repository access. After project completion, there's potential for an ongoing maintenance contract as we scale to 100+ users.
Ideal background: React/TypeScript senior with hands-on WebRTC and real-time audio, comfortable steering AI-generated code to production quality. If you can start immediately and wrap within 3–4 weeks, let's get this live.
To stand out, please briefly mention:
1. Your most complex WebRTC/real-time audio project
2. Experience with Supabase Edge Functions or similar serverless
3. Your approach to debugging async audio/transcription timing issues
4. Available hours per week over the next month
Timeline: Start this week, MVP stable within 2 weeks, full completion in 3-4 weeks.
The platform conducts voice interviews using OpenAI's Realtime API, capturing and analysing responses in real-time. Four Supabase Edge Functions handle token generation, interview processing, and notifications—these may need debugging alongside the WebRTC fixes. Audio streams at PCM16/24kHz. The React app connects to OpenAI through a Supabase Edge Function relay for security. Session management and ephemeral tokens are already implemented but may need refinement during WebRTC debugging.
The first milestone is, therefore, a rock-solid WebRTC fix. I need those dropped-connection edge cases handled, with graceful reconnection and accurate transcription timestamps, all verified in the live production environment we'll ultimately ship on Vercel.
Once the stream is stable, the scope widens:
- Finish the AI response pipeline so answers come back in real time
- Integrate Stripe for subscription billing (monthly & annual plans already defined)
- Build an admin dashboard for the interview template and user management
- Add five missing pages—results, settings, library, plus two small utility screens—styled to match existing Tailwind components
- Implement global error handling, sensible API rate-limiting, and security hardening
- Optimise bundle size and lighthouse performance before final deploy
I will provide full repository access, existing Supabase tables with 13 configured tables and RLS policies, and the existing UI/component structure using shadcn/ui. Please keep the current structure; the goal is completion, not a rewrite. Clear inline docs and a short handover guide are part of the final deliverable so my team can maintain it.
NDA required before repository access. After project completion, there's potential for an ongoing maintenance contract as we scale to 100+ users.
Ideal background: React/TypeScript senior with hands-on WebRTC and real-time audio, comfortable steering AI-generated code to production quality. If you can start immediately and wrap within 3–4 weeks, let's get this live.
To stand out, please briefly mention:
1. Your most complex WebRTC/real-time audio project
2. Experience with Supabase Edge Functions or similar serverless
3. Your approach to debugging async audio/transcription timing issues
4. Available hours per week over the next month
Timeline: Start this week, MVP stable within 2 weeks, full completion in 3-4 weeks.