Worked on the pytorch/pytorch repository to enhance cross-platform GPU support by enabling HIPification of CUDA code. Focused on adding mappings for cudaMallocAsync and cudaFreeAsync within the cuda_to_hip_mappings, this contribution laid the groundwork for smoother migration from CUDA to HIP. Leveraging expertise in CUDA, HIP, and GPU programming, the developer implemented these changes using Python, ensuring that PyTorch’s GPU backend could better support multiple platforms. The work was feature-driven, with no major bug fixes, and addressed compatibility and performance needs for future development. This targeted update supports PyTorch’s ongoing strategy for broader GPU programming flexibility and maintainability.