AI Voice Agent Service Development

Senior AI Voice Engineer - Pipecat, Twilio & Python We are looking for a senior engineer with hands-on experience building real-time AI voice systems to develop a standalone voice-agent service for an existing SaaS platform. This is not a basic chatbot or prompt-engineering project. The core challenge is building a reliable, production-quality real-time voice service that can conduct autonomous phone calls, navigate real-world telephony conditions, interact with external APIs during calls, and return structured results to our application. The initial use case involves workflow-driven outbound calls for a professional-services platform. We have a detailed functional specification that will be shared with shortlisted candidates rather than posted publicly. Technology Our anticipated stack for this project includes: * Python * Pipecat * FastAPI * Twilio / Twilio Media Streams * OpenAI Realtime API and/or OpenAI models * Pluggable speech-to-text and text-to-speech providers * REST APIs / WebSockets * Docker We are specifically interested in candidates who have worked with real-time audio and telephony systems- not simply developers who have integrated an LLM into a web application. What You Will Build You will own development of a standalone voice-agent service that communicates with our existing platform through a defined API. At a high level, the service will need to: * Initiate and manage AI-driven phone calls through Twilio * Run low-latency real-time voice conversations * Handle natural interruptions and turn-taking * Navigate IVRs and send DTMF input * Deal intelligently with holds, silence, voicemail, transfers, disconnects, and other real-world call states * Allow the AI agent to invoke application functions during a conversation * Capture structured information gathered during calls * Support configurable agent behaviors, prompts, voices, and languages * Return transcripts, call outcomes, structured results, operational metrics, and related data to our API * Be designed so additional voice-agent use cases can be added without rebuilding the underlying service Reliability and architecture matter as much as getting the demo working. What We Will Provide We will provide shortlisted/hired developers with: * A detailed project specification and acceptance criteria * API documentation * A mock API/server for development * Synthetic test data * Agent prompts and expected structured outputs * Development credentials for required third-party services * Access to our engineering team for integration questions Development will occur in a standalone repository. You will not need access to our existing production application or production customer data. Expected Deliverables The engagement will include: * Production-quality Python service * Configurable agent architecture * Dockerized development/deployment environment * Automated tests * End-to-end call testing * Technical and deployment documentation * Engineering handoff We expect the project to be broken into defined milestones. Who We're Looking For Strong candidates will have significant experience with several of the following: * Pipecat or comparable real-time voice-agent frameworks * Twilio Voice and Media Streams * WebSocket-based real-time audio * OpenAI Realtime API * STT/TTS pipelines such as Deepgram, Cartesia, ElevenLabs, or similar * Voice activity detection and interruption/barge-in handling * IVR/DTMF automation * Async Python * FastAPI * Function/tool calling * Low-latency distributed systems * Production monitoring, retry, and failure handling Experience building production telephone agents is significantly more valuable to us than general LLM or chatbot experience. Application Please do not send a generic AI-generated proposal. Instead, tell us about the most technically challenging **real-time voice or telephony system you personally built**. Explain what you were responsible for and which technologies you used. If your experience is primarily with platforms such as Vapi, Retell, Bland, Synthflow, or similar managed voice products, please say so. That experience is useful, but this project requires building more of the underlying voice infrastructure directly. Shortlisted candidates will receive the full technical specification before we finalize scope, milestones, and pricing.