Overview
Develop comprehensive performance models and projections for our architecture across varying workloads and configurations. Profile and analyze deep learning workloads on our hardware to identify micro-architectural bottlenecks and influence optimization opportunities. Drive hardware/software co-optimization by identifying where architectural features can unlock performance improvements.