Worked on the intel/intel-xpu-backend-for-triton repository to enhance NVIDIA backend fidelity by refactoring the lowering of ttn::ClusterCTAIdOp. Focused on replacing inline PTX assembly with a sequence of NVVM operations, this approach preserved more semantic information at the LLVM level and reduced reliance on bespoke assembly fragments. The work, implemented in C++ and MLIR, improved code maintainability and positioned the backend for future performance optimizations. No major bug fixes were addressed during this period, as efforts centered on feature delivery, backend stabilization, and enabling more robust low-level optimizations for GPU programming within the compiler development workflow.