Worked on the pymc-devs/pytensor repository to enhance the robustness and correctness of core tensor operations, focusing on the AdvancedSubtensor component. Addressed a key bug by refactoring AdvancedSubtensor to treat newaxis as an explicit DimShuffle and to encode slices internally, ensuring that only numeric inputs are accepted. This approach centralized slice encoding logic, simplifying future maintenance and improving error messaging for developers. Collaborated closely with another contributor to deliver these changes. Utilized Python and applied advanced indexing, code refactoring, and tensor manipulation skills to reduce runtime errors and prevent misuse, ultimately strengthening input validation across the codebase.