AI Voice Engineer — Real-Time Conversational AI + Telephony MVP

I’m looking for an experienced AI Voice / Real-Time Conversational AI engineer to help build the MVP of a confidential voice-based platform. The core product involves users calling a single inbound phone number, being authenticated/identified, and then being automatically routed into an individual real-time AI voice session. The system needs to support natural, low-latency, two-way conversations and maintain a persistent user-to-AI-assistant relationship across calls. This is not a simple chatbot project. I’m specifically looking for someone who has experience building production-quality voice AI + telephony infrastructure. Core MVP flow Caller → Phone number → Authentication/payment → User identification → Assigned AI companion → Real-time voice conversation → Call ends → Usage recorded → Caller can call again The initial MVP should be intentionally lean. I do not want unnecessary features or an over-engineered platform. Technologies / experience we’re looking for Strong experience with several of the following: * Twilio * SIP / WebRTC * OpenAI Realtime API or comparable real-time voice AI * ElevenLabs or comparable voice technology * Real-time audio streaming * Conversational AI * Voice activity detection / interruption handling * Low-latency voice pipelines * Node.js and/or Python * REST APIs * Webhooks * Database architecture * Authentication * Payment integration * Usage/minute tracking * Cloud deployment * Scaling concurrent voice sessions The architecture should eventually support * One primary inbound phone number * Large numbers of registered users * Persistent user accounts * Individual AI companion assignments * Multiple simultaneous conversations * Conversation/session management * Usage tracking * Payment/authorization * Ability to scale from hundreds to thousands of concurrent calls * Monitoring and analytics Important: I am looking for someone who understands that the major scalability considerations are concurrent calls, telephony capacity, real-time AI capacity, latency, infrastructure and cost per minute, rather than simply purchasing additional phone numbers. What I need from you Before hiring, I want to understand how you would architect this system. Please answer these questions: 1. Have you personally built a production voice-AI application involving real-time phone calls? Please provide an example. 2. What telephony provider and voice-AI architecture would you recommend for this project and why? 3. How would you handle thousands of users calling the same number while maintaining separate AI sessions? 4. How would you maintain a persistent relationship between: Caller → User ID → Assigned AI companion → Companion configuration 5. How would you handle interruptions, latency, dropped calls and reconnects? 6. How would you design the architecture so it can scale from: 100 → 1,000 → 10,000+ concurrent calls? 7. What would you estimate the infrastructure/AI cost per minute to be at those different scales? 8. What would you build during the first 2–4 weeks, and what would you deliberately leave out of the MVP? MVP Deliverable The first milestone should demonstrate a complete working flow: Caller ↓ Inbound phone number ↓ Authentication/payment ↓ User identified ↓ Assigned AI voice ↓ Natural two-way conversation ↓ Call ends ↓ Usage recorded ↓ Same user can call again and reach their assigned AI I want to be able to actually call the system and test it, not simply receive source code. Ideal candidate You are a strong fit if you have: * Built voice AI products before * Worked with Twilio/SIP/WebRTC * Worked with OpenAI Realtime or similar technology * Experience with streaming audio * Experience designing for concurrency * Strong backend/API skills * Understand production deployments * Can explain technical decisions in plain English * Can demonstrate previous work Please do not apply if your experience is primarily ChatGPT integrations, prompt engineering, or basic text-based AI applications.