Worked on the nod-ai/iree-amd-aie repository to deliver a targeted vectorization enhancement for the AMD AIE backend. Developed a new lowering pattern in C++ and MLIR that transforms vector.multi_reduction operations into vector.reduction, implemented within the AMDAIEVectorizationPass using the populateVectorMultiReductionLoweringPatterns method. This approach enables subsequent vectorization stages and lays the groundwork for improved throughput and consistent code generation. Added comprehensive regression testing with multi_reduction_to_reduction.mlir to validate the transformation across various reduction operations and data types. The work focused on compiler development, low-level optimization, and vectorization, supporting future enhancements in the codebase.