Projects
A collection of my work spanning AI/ML, web development, and interactive experiences.
BreakingWeb
Constructing challenging browser-use tasks by controlled environment interventions — matched clean/intervention task pairs that measure the cost of each intervention and isolate the primary recovery demand on web agents.
AI AgentsBenchmarkingWeb AgentsEvaluation
ComponentBench
A benchmark diagnosing where computer-use agents fail inside modern UI components — 2,900+ tasks across 97 component types, with human baselines and per-family failure analysis.
Computer-Use AgentsBenchmarkingUI ComponentsEvaluation
Interested in working together?
Get in Touch