Leads a group developing high-throughput virtual environment simulators like the Madrona Engine and BPS3D for AI agent training
Pioneers new applications using massive video datasets, including analysis of 250,000+ hours of cable news and creation of interactive, realistic tennis player avatars from broadcast footage
Built scalable systems for video database processing and querying, improved optimizing compilers for image processing, and advanced DNN inference techniques for video
Develops interactive, expert-in-the-loop ML systems for generative AI workflows and rapid model training/validation
Performs path tracing on terabyte-scale scenes using thousands of AWS Lambda functions