Developed and delivered a major feature for the EnzymeAD/Enzyme-JAX repository, enabling partial symmetry detection in tensor operations such as transpose and dot_general. This work involved implementing generation and propagation logic for symmetry, optimizing performance, and expanding support for symmetry-aware transformations in high-performance machine learning workloads. The developer refactored symmetry generation, extended handling for equality cases, and introduced optimizations like n-dimensional transpose removal and IR annotation recognition. Using C++, MLIR, and advanced algorithm optimization, they addressed correctness issues, improved aliasing checks, and expanded test coverage, resulting in more reliable and maintainable tensor analysis within the EnzymeAD codebase.