Contributed to the Xilinx/onnx-mlir and onnx/onnx-mlir repositories by developing four new features over two months, focusing on compiler development and low-level optimization. Implemented BitwiseNot ONNX operation support by lowering it to Krnl using XOR logic, and updated documentation to guide users. Expanded the ONNX-to-Krnl lowering path with Mean-Variance Normalization, generalized window function support for Hamming and Blackman windows, and enabled RandomUniform tensor generation. These enhancements, written in C++ and MLIR, improved model normalization, operator coverage, and runtime capabilities, aligning with project goals to increase portability, performance, and deployment readiness for machine learning frameworks.