Worked on the ggml-org/llama.cpp repository to deliver Intel Xe1 Vulkan Cooperative Matrix Support, enabling accelerated matrix operations on Intel Xe-LPG Plus GPUs. The implementation involved updating architecture enums, removing legacy driver workarounds, and introducing Windows driver checks to improve deployment reliability. Focused on maintainability by addressing code quality issues such as bf16 handling and eliminating trailing whitespace. Utilized C++ development skills alongside GPU programming and Vulkan to lay the foundation for broader hardware acceleration. The work improved performance potential on supported GPUs and established a clearer path for future hardware targets, reflecting a methodical and forward-looking engineering approach.