Worked on enhancing the PyTorch repository by improving documentation for the CUDA/MTIA device API, focusing on adding comprehensive docstrings to the check_error and is_bf16_supported functions. Leveraged Python and deep familiarity with PyTorch codebase conventions to align new documentation with established standards, ensuring clarity and consistency. The approach emphasized clear guidance on device-related error handling and BF16 support, directly addressing common developer pain points. This work aimed to streamline onboarding for new contributors and reduce ambiguity in API usage, ultimately supporting smoother integration and fewer support requests. The contribution centered on API documentation, Python, and open-source collaboration workflows.