Worked on enhancing cross-platform compatibility in the pytorch/pytorch repository by implementing a CUDA to HIP host function mapping. Focused on enabling host function type conversions, the work involved updating the cuda_to_hip_mappings table to support mapping from cudaHostFn_t to hipHostFn_t. This technical approach streamlines the process of porting PyTorch workloads that rely on host function callbacks, reducing manual effort for developers targeting both CUDA and HIP backends. Utilized Python alongside GPU programming skills and expertise in CUDA and HIP interoperability. The contribution lays foundational support for broader portability of host-function callbacks within the PyTorch ecosystem, though no bugs were addressed.