During June 2026, contributed to the ROCm/aiter repository by implementing GELU_BIAS epilogue support for hipb_mm matrix multiplication. This work involved updating the Python interface, C++ bindings, and CUDA kernels to handle the new epilogue type, ensuring seamless integration across both frontend and backend components. Comprehensive tests were added to verify the correct functionality of GELU_BIAS and to enforce bias provisioning requirements. The changes expanded model expressiveness for neural networks and improved the robustness of the ROCm/aiter codebase. The work demonstrated proficiency in backend development, with a focus on C++, CUDA, and Python, and emphasized thorough testing practices.