Worked on the NVIDIA/cuda-python repository to modernize CI/CD workflows and enhance CUDA kernel interface performance. Focused on transitioning CI configuration from JSON to YAML, updating workflows to define CUDA build versions and test matrices, and removing outdated files to improve maintainability and readability. Refactored kernel argument type handling by replacing isinstance checks with explicit type comparisons, which improved performance and ensured correct handling of Python boolean objects. Leveraged Python, YAML, and Cython to streamline CI processes and kernel preparation logic. These changes reduced configuration complexity, standardized commit practices, and made onboarding and pipeline updates more efficient for future contributors.