Worked on the llvm/torch-mlir repository to expand arithmetic lowering from Torch to MLIR by implementing robust conversion patterns for several Aten operations. Leveraging C++ and MLIR, introduced unified templates to handle AtenSubOp, AtenMulOp, and AtenDivOp, as well as a dedicated conversion for AtenNegFloatOp to arith::NegFOp. This approach increased automatic coverage of arithmetic operations, reduced the need for manual op-by-op lowering, and improved consistency across the conversion path. The work focused on maintainable, template-based patterns, enabling broader model deployment and enhancing downstream optimization opportunities within the compiler design and C++ development domains.