Worked on the kubernetes/autoscaler repository to enhance observability and reliability of cluster scaling operations. Developed new metrics in Go to monitor both scale-down and scale-up processes, including tracking the maximum evaluation time for skipped nodes and integrating a metric observation interface into the ClusterStateRegistry. Focused on backend development and instrumentation, these changes improved visibility into autoscaler decisions and enabled data-driven tuning for cost optimization. Comprehensive unit tests were added to ensure accurate metric generation during scale-up failures, supporting faster diagnosis and resolution. The work emphasized code quality, best practices in metrics, and robust unit testing within a Kubernetes environment.