Background
Director of UC Berkeley's Sky Computing Lab; turned academic open source (Spark, Ray) into Databricks and Anyscale. He was Wei-Lin Chiang's PhD advisor, connected Angelopoulos to the project, and served as Arena's interim head of machine learning (Felicis).
What they run now
Board-level strategy and founder mentoring at Arena; publicly explains Arena as AI's neutral referee and highlights its move to simulated-preference models and the Max routing product.
Career
- 2025 – presentCo-founder & Chairman, Arena
- 2013 – presentCo-founder & Executive Chairman, Databricks
- presentProfessor; director, Sky Computing Lab, UC Berkeley
- presentCo-founder, Anyscale
On the record
Humans stay essential for deciding what to solve and verifying results; AI mainly automates the middle step, and experts get the most out of it. source
Expects real value but a correction, as with any new technology that overshoots. source
Describes Arena building models that simulate user preferences for automated evaluation at scale, powering the Max routing product. source