Junqi Ye developed GPU-to-job metrics collection and observability features for the aws-samples/awsome-distributed-training repository, focusing on enhancing traceability and performance analysis in Slurm-managed environments. By implementing prolog and epilog bash scripts, Junqi enabled end-to-end mapping of GPUs to Slurm jobs, allowing for per-job metrics collection. The solution integrated DCGM and OTEL-based exporters, adding Slurm job IDs as metric attributes to improve monitoring granularity. Junqi also ensured script persistence across container restarts through directory mapping in Docker configurations. This work demonstrated depth in containerization, DevOps, and monitoring, providing a robust foundation for richer observability in distributed training workflows.