Worked on the aws/amazon-vpc-cni-k8s repository to improve the accuracy and utility of Prometheus metrics for VPC networking operations. Refactored the metrics system by converting gauges to counters, ensuring that metrics now represent total occurrences of AWS API errors, pod ENI errors, and force removals of ENIs and IPs rather than just current state snapshots. This change enhanced observability and reliability, allowing dashboards and alerts to better support incident response and capacity planning. Utilized Go and Prometheus instrumentation to implement these updates, maintaining code quality through targeted changes that minimized risk and facilitated future evolution of the metrics module.