Jingning Tang developed and enhanced benchmarking and usability features across the ROCm/triton, ROCm/aiter, and intel-xpu-backend-for-triton repositories. Using Python, C++, and CUDA, Jingning refactored the plot layout command-line interface in ROCm/triton to improve argument parsing and maintainability. In ROCm/aiter, Jingning created a GEMM A8W8 benchmarking script that generates input data, runs multi-shape benchmarks, and parses results for performance analysis. For intel-xpu-backend-for-triton, Jingning expanded the MI300 benchmarking suite to support additional data types and cross-architecture compatibility, addressing dtype mismatches and optimizing block sizes to enable broader hardware coverage and more robust performance evaluation.