I lead Applied AI Research at GitHub, working on the AI coding systems behind GitHub Copilot.
My team builds the foundations that make AI agents practical at scale: evaluation systems that catch quality regressions before they reach users and surface where agentic harnesses have room to hillclimb, multi-agent architectures that improve task completion while keeping token cost in check, and memory and context platforms that cut the ramp-up cost of every new session by helping GitHub Copilot understand developers, repositories, and workflows over time.
I work closely with engineers, researchers, and product teams to turn advances in AI into reliable products used by millions of developers.
Before GitHub, I worked in both industry and academia, building machine learning systems and teams at Google, Allianz, INRIA, HIIT, and Aalto University.
I believe the best innovations come from combining rigorous thinking, practical execution, and a genuine curiosity about how things work.
