Voice AI Engineer (Real-Time Phone Agents), AI Healthcare

Long-term, full-time project. Individuals only. Linear Health builds AI that does the back office work US clinics can't keep up with: referrals, prior authorizations, scheduling, patient outreach, and voice agents that call and answer patients. Our customers are community health centers and specialty practices. WHY THIS ROLE EXISTS Our voice agents talk to real patients every day, in English, Spanish and other languages. A good demo is easy. A voice agent that handles interruptions, silence, bad phone lines and real patient records, at a cost per minute that makes sense, is hard. We need one person who owns all of our voice agents end to end. WHAT YOU'LL OWN • Every voice agent across every customer, from the phone line to the patient record • The phone and audio layer: speech recognition, turn-taking, interruptions, silence and latency • The agent layer: prompts, tools, conversation state and confirmations before anything is written to a patient record • Cost per call minute, measured per turn and brought down without hurting quality • Testing and monitoring, so problems show up before a clinic tells us • Custom build a voice ai monitoring system too if you like YOU'LL FIT IF YOU • Have built real-time voice agents that took real calls in production, not demos • Have worked underneath the platform: LiveKit, Pipecat, Twilio, SIP, speech recognition or turn detection. Vapi or Retell alone is not enough and n8n is not real engineering work • Are strong in Python and FastAPI • Have 2+ years of full-time work owning a production system • Can trace a bad call from the logs and prove the cause before changing anything • Use AI coding tools every day and know where they go wrong NICE TO HAVE • Healthcare, or other regulated phone calls like collections or insurance • ElevenLabs Agents • Real numbers on latency or cost per minute that you improved HOW WE WORK • Full time and exclusive. Individuals only, no agencies or teams, and no other clients once you start • Minimum 6 hours a day of overlap with US Central time, Monday to Friday • You'll complete our HIPAA and security training before you start • Small, flat team. You own voice from day one HOW TO APPLY (PLEASE READ CAREFULLY) Start your proposal with the word "Signal". We automatically reject any proposal that uses em dashes (the long dash), so please write it yourself. Then read the attached task and record a Loom of no more than 3 minutes, with your camera on, walking us through how you'd approach it. Put the link in your proposal. Don't build anything. We want to hear how you think. We're not looking for a perfect answer. Please don't read from a script or send an AI-written answer. If we detect one, the application is rejected.