Worked on the llvm/clangir repository to deliver a performance-focused optimization in the InstCombine pass, targeting ceiling division by powers of two. The approach transformed a common right-shift and conditional-add pattern into a single addition followed by a right shift, provided no intermediate overflow occurs, reducing instruction count and improving code generation. Comprehensive tests were added to ensure correctness and handle edge cases, reinforcing safe transformation patterns. The work demonstrated strong skills in compiler optimization, instruction combining, and low-level systems programming using C++ and LLVM IR. No major bugs were reported, and the changes contributed to broader optimization coverage in ClangIR.