Worked on the google-research/timesfm repository to deliver a feature that allows users to specify a configurable local directory for model snapshot downloads. This addition gives users greater control over storage location, supporting improved data management and deployment flexibility. The implementation involved integrating new path configuration options into the existing model snapshot workflow, ensuring robust path handling and maintaining backward compatibility. All changes were tracked with explicit commit-level traceability for auditability. Leveraging Python and applying skills in data management and machine learning, the work focused on enhancing reproducibility and supporting diverse user workflows without introducing regressions or breaking existing functionality.