Worked on the espressif/llvm-project repository to deliver a targeted compiler optimization feature focused on SIMD performance. Developed enhancements in the InstCombine pass to widen the select instruction’s width after a comparison, enabling the generation of Max and Min intrinsics. This approach improved code generation by allowing better handling of type mismatches in select operations, resulting in broader intrinsic support for vectorized workloads. Utilized C++ and leveraged expertise in LLVM IR, instruction selection, and compiler optimization techniques. The work demonstrated a disciplined approach to back-end compiler engineering, addressing specific code generation challenges without introducing new bug fixes during the period.