About the role
Aurora's core promise is an AI agent that can call a business, hold a natural conversation, and come back with a verified answer. You'll own the systems that make that conversation possible — speech recognition, turn-taking, latency budgets, and the glue between telephony and our reasoning models.
This is a founding role. You'll set the technical direction for voice at Aurora, make the build-vs-buy calls, and see your work dial real businesses within days of merging.
What you'll do
- Design and own the real-time voice pipeline — STT, TTS, turn-taking, and barge-in handling over live phone calls.
- Drive end-to-end latency down to conversational levels and keep it there as we scale.
- Build evaluation harnesses that measure whether calls actually accomplish the user's task.
- Work directly with founders on model selection, prompting strategy, and call-flow design.
What we're looking for
- Strong systems engineering background — you've built and operated low-latency, real-time services in production.
- Hands-on experience with speech or audio pipelines (STT/TTS, WebRTC, SIP, or similar).
- Comfort owning ambiguous problems end to end with minimal process.
- Bias toward shipping: prototypes this week beat proposals next quarter.
Nice to have
- Experience with telephony infrastructure (Twilio, LiveKit, or raw SIP).
- Prior work with LLM agents, tool use, or conversation state machines.
- Early-startup experience — you've been employee #1–10 somewhere.