Senior AI Engineer — Voice
Mira Mace, San Francisco
What you will do
- Turn goals into a roadmap. Take a goal like "more of the people who sign up get through their first call" or "more insurer calls end with the answer we needed," break it into the problems behind it, size them, and decide what to build first. You write the plan, defend the tradeoffs, and change it when the data says to.
- Own the voice pipeline. Build and operate the real-time stack — telephony, streaming speech-to-text, LLM orchestration, text-to-speech, endpointing, barge-in, and the latency budget that keeps a conversation feeling human.
- Drive call quality relentlessly. Define what a good call is, measure it, and move the number. Build the evaluation harness — replayable audio and transcripts, regression suites, per-turn scoring — so quality changes are provable rather than anecdotal.
- Handle the real world. Automated phone menus and DTMF, holds and transfers, voicemail, dropped calls and retries, noisy lines, accents, interruptions, and speaking clearly with an older audience.
- Scale calling volume reliably. Make the system hold up across thousands of calls a day, with clean handoffs to a human when the agent should stop.