Founding AI Engineer (Speech / Voice AI)

Most voice AI works in demos. It breaks in production.

Real conversations involve interruptions, dialect switching, background noise, latency spikes, and compliance constraints. Most teams aren’t solving for that.

This YC startup is.

They’ve just come out of YC, grown from ~$5K to ~$50K revenue in a very short window, and already have customers live in production. They’re now building the infrastructure and evaluation layer behind real-world voice systems - not just agents, but how those systems are monitored, improved, and trusted.

If you care about how voice actually works at scale, this is where the real problems are.

The Role

This is a Founding AI Engineer position.

You’ll work directly with the founders and own the AI direction as an individual contributor. No people management. High autonomy.

You won’t be handed tickets. You’ll define:

* What gets built

* How systems are evaluated

* What “good” voice interaction actually looks like

What You’ll Work On

* Designing evaluation systems beyond WER/CER

* Improving naturalness and flow in real-time conversations

* Solving latency and streaming challenges

* Building speech systems that handle real-world complexity

* Creating feedback loops to continuously improve voice agents

What They’re Looking For

You’ve likely worked on:

* Speech / audio / conversational AI systems

* ASR / TTS / speech-to-speech pipelines

* Real-time or low-latency ML systems

And importantly:

* You’ve worked at the model level (not just APIs)

* You’ve deployed systems into production environments

Backgrounds that fit well:

* Research Engineer or Applied Scientist (speech / audio)

* Experience in a voice AI startup or relevant Big Tech team

* PhD is a plus, but not required

Why This Role

Most roles force a trade-off:

* Research vs product

* Depth vs ownership

This doesn’t.

You get:

* Real research problems (voice is still unsolved)

* Direct product impact

* Founding-level ownership without management overhead

If you’ve been a small part of a larger system, this is a chance to define the system itself.

Process

* Intro with founders

* Technical conversations

* Short paid trial (1-2 weeks)