Peggy Tang worked on the NVIDIA/cuda-samples repository, focusing on resolving CUDA runtime constraint issues to improve kernel launch reliability and debug-mode stability. She addressed a bug by increasing the pending kernel launch limit in cdpAdvancedQuicksort.cu and enforcing a per-thread register usage cap in the conjugateGradientMultiBlockCG sample’s CMake configuration. Using C++ and CMake, Peggy applied performance tuning and debugging skills to align sample applications with GPU hardware constraints, reducing the risk of launch failures with larger workloads. Her targeted changes enhanced maintainability and future-proofing of the codebase, demonstrating depth in understanding both CUDA runtime behavior and sample configuration.