The Race to Efficient Compute: From Workload to Silicon

Nir Sharvit | Director of Edge Compute & Physical AI

This session explores a workload-driven approach to designing efficient edge compute architectures, connecting AI workload characteristics to memory, compute, system architecture, compiler mapping, and software orchestration. It will examine how software-first co-design, digital twins, and behavioral SoC models can help engineers evaluate architectural tradeoffs and optimize systems before silicon is available.