Worked on stabilizing the Elastic Job Scale-Down workflow in the kubernetes-sigs/kueue repository, addressing reliability issues in Kubernetes controllers using Go. Focused on backend development to resolve a bug where stale reclaimable pod counts could cause the reconciler to stall, introducing logic to clamp reclaimable counts to pod set sizes and updating admission webhooks to tolerate stale data during scale-down transitions. Expanded test coverage to include edge cases, ensuring the fix was robust and reproducible. The work improved the predictability and reliability of elastic job scale-down operations, reducing the risk of workflow stalls and premature quota denials in production environments.