Teaching computers to help people.
Principal Software Engineer at Weights & Biases, now part of CoreWeave. I build CoreWeave ARIA, a research agent for ML teams. Before that I led Weave, our open-source toolkit for tracing and evaluating AI applications, and earlier ran ML platform products at Twitter and search at Workday.
Play talk
An AI research agent embedded in W&B. Ask questions of your experiments, runs, artifacts and charts in plain language. The goal is an always-on collaborator that proposes hypotheses, launches experiments on cloud GPUs, and iterates on your behalf.
Observability and evaluation for production agents. Trace, evaluate and improve AI applications, with sessions, turns and tool calls treated as first-class concepts.
Set in Newsreader and Hanken Grotesk.