During August 2025, Rui Lin enhanced TPU job tracking and resource allocation in the apple/axlearn repository by developing a feature that appends project-id and num-replicas labels to TPU job configurations. Leveraging Python, GCP, and cloud computing expertise, Rui implemented labeling-based instrumentation to improve observability and traceability of TPU workloads. This approach laid the foundation for more effective cost allocation, auditing, and SLA reliability by enabling granular tracking of resource usage. The work focused on robust unit testing to ensure reliability and maintainability, resulting in a well-integrated solution that addressed the need for better monitoring and planning of TPU resources.