Worked on the vllm-project/vllm-omni repository, focusing on codebase cleanup to enhance maintainability and streamline future development. The primary contribution involved removing the unused video.py asset, which previously handled audio extraction from video files but was no longer needed. This targeted removal reduced repository clutter and minimized maintenance risks, supporting a cleaner development environment. Emphasizing clean code and effective code maintenance practices, the work improved overall code hygiene and facilitated easier onboarding for new contributors. The process was managed using Python and careful asset management, with attention to maintaining CI workflow integrity by skipping unnecessary CI runs during cleanup.