Worked on the Xilinx/llvm-project repository to deliver an API enhancement within MLIR Affine Dialect Utilities, focusing on memory allocation normalization flows. Extended the affine::normalizeMemRef utility to return a FailureOr containing the normalized memref.alloca or memref.alloc operation, rather than a simple LogicalResult, thereby improving downstream usability for transformation passes. This change enables more streamlined and informative memory allocation workflows, supporting advanced compiler development and IR manipulation tasks. The work was implemented in C++ and required a strong understanding of low-level optimization techniques, contributing to the overall robustness and flexibility of the MLIR Affine Dialect’s API design.