Worked on the google/orbax repository to enhance observability and monitoring for checkpointing workflows. Overhauled the metrics system by implementing Prometheus-based metrics emission and introducing a Prometheus monitoring class with environment-driven telemetry initialization, replacing legacy Google Monitoring components to streamline maintenance. Delivered a feature to track storage types used during checkpoint load operations, enabling improved data management and capacity planning. Focused on backend development using Python, Prometheus, and telemetry instrumentation, ensuring low-overhead integration without impacting API stability. The work emphasized maintainability and actionable diagnostics, supporting faster incident response and data-driven optimization for production environments handling checkpoint data.