Cen Zhao developed a CUDA Graphs Benchmarking Mode for the facebookresearch/param repository, introducing a standardized workflow for GPU performance analysis. Using Python and leveraging CUDA and distributed systems expertise, Cen implemented a new argument to enable graph-based benchmarking, which captures and replays CUDA operations through a defined sequence of warmup, graph capture, and replay. The solution included device gating for CUDA and ROCm compatibility and addressed asynchronous operation handling to ensure reliable measurements. This work established a reproducible benchmarking process, enabling cross-hardware performance comparisons and supporting faster optimization cycles, reflecting a deep understanding of performance optimization and benchmarking challenges.