Worked on the NVIDIA/cuda-python repository to enhance both documentation and cross-platform support for CUDA memory management features. Focused initially on improving API clarity by updating type annotations and attribute descriptions for VirtualMemoryResourceOptions, which reduced ambiguity and aligned documentation with code semantics. Subsequently, implemented Windows support for CUDA-enabled VirtualMemoryResource by introducing runtime checks for CUDA VMM API availability, replacing previous hard-coded disablement. This approach improved platform parity and developer productivity by ensuring features are enabled based on actual API presence. The work leveraged Python, CUDA, and memory management expertise, contributing to safer, more maintainable bindings and a better developer experience.