Worked on the NVIDIA/cuda-python repository to deliver enhancements to the CuFile I/O API, focusing on handwritten bindings for cuFileRead and cuFileWrite. Leveraging Python and Cython, developed a comprehensive testing suite that included data integrity checks, internal validation, and statistics collection to ensure robust and reliable read/write operations. Added targeted test coverage for CUDA 13.0, validating compatibility and performance of CuFile I/O operations. The work emphasized thorough validation across all I/O paths, resulting in improved reliability and maintainability of the API. This approach demonstrated depth in CUDA integration, Python development, and systematic testing for high-performance data workflows.