UW student Zhihan Zhang wins the Madrona Prize for shrinking AI models onto small chips

A University of Washington project that uses AI agents to fit models onto small, low-power chips won the Madrona Prize at the Allen School Research Showcase on Thursday, GeekWire reported. Madrona has given the prize since 2006 to student projects that combine research with commercial potential. PhD student Zhihan Zhang's winning work tests whether agents can shrink models and tune device software for chips with tight limits on memory, power and heat.

Leading models failed at the task without feedback from real hardware. The team let the agent load its work onto hardware, measure the result and try again. With that loop it beat human experts within seven rounds. One test was an elk-detection camera trap that had to run on a small solar-powered device.