Worked on reorganizing the AI-Hypercomputer/tpu-recipes repository to improve clarity and maintainability for users and contributors. The main focus was archiving outdated PyTorch/XLA training recipes by moving them into a dedicated archive directory and updating associated READMEs to direct users toward actively maintained MaxText-based alternatives. This restructuring effort, implemented using Markdown and Shell scripting, streamlined the repository’s structure and documentation, reducing confusion around recommended workflows. By consolidating legacy content and clarifying onboarding materials, the work laid a foundation for future refactors and made it easier for new contributors to navigate and adopt current best practices.