Worked on the harvard-edge/cs249r_book repository, focusing on refactoring the log_softmax function to enhance code clarity and maintainability. Applied Python programming and numerical methods to simplify logic, remove redundant code, and establish reusable patterns for future loss computations in machine learning workflows. The approach involved renaming and reusing variables to reduce cognitive overhead and minimize the risk of regressions, supporting safer extension across related modules. Emphasized clear naming conventions, disciplined refactoring, and clean commit practices throughout the process. This work laid a foundation for easier testing and future feature development, reflecting a methodical and quality-driven engineering approach.