Worked on the googleapis/python-aiplatform repository to refactor TensorBoard plugin name resolution, focusing on improving code maintainability and reducing internal dependencies. The approach involved replacing direct imports from metadata modules with the use of public plugin_name attributes on plugin classes, which enhances robustness and decouples the integration from internal metadata changes. This targeted update, implemented in Python and leveraging skills in code maintenance and refactoring, supports safer future updates to plugin names and simplifies ongoing maintenance. The work addressed a specific area of the codebase, resulting in a more stable TensorBoard integration and reducing the risk of future breaking changes.