What I'm doing now

Building

Leading AI engineering at Invoca: conversational AI that revenue teams rely on, at enterprise scale. Deep in the practical realities of voice, agents, models, and evaluation.

Writing

Publishing essays here on conversational AI and LLMs in production, starting with the story of building Nebula and why human conversations are still the hardest thing language models do.

Thinking about

Evaluation as the center of AI engineering: how you measure whether a system actually did its job, and what happens to conversation itself as agents start talking to agents.