Own the end-to-end voice AI pipeline, optimizing STT, LLM, and TTS latency for sub-second conversational responses. Responsibilities include provider evaluation, abstraction-layer development, turn detection, interruption handling, filler audio, retries, observability, tracing, and build-versus-buy decisions. The role requires hands-on production experience with cascaded voice systems, strong systems fundamentals, on-call ownership, and comfort solving ambiguous problems with real production data.