Worked on the llvm/torch-mlir repository to enhance sparse tensor workflows in MLIR, focusing on enabling the Sparse Tensor Dialect across all MLIR rewrites. Addressed a dependency issue by registering the sparse tensor dialect, allowing users to perform end-to-end sparse-tensor rewrites even when building without StableHLO. Developed and later streamlined an example PyTorch-MLIR compiler, MPACT, to demonstrate sparsity propagation, improving documentation clarity and reducing maintenance overhead. Emphasized build stability and user experience by ensuring core MLIR integration supported sparse dialects. Utilized C++, MLIR, and PyTorch, with attention to technical writing and documentation hygiene throughout the development process.