Worked on the kubernetes/autoscaler repository to address a stability issue in the scheduler framework related to PersistentVolumeClaims. Focused on backend development using Go and Kubernetes, the work involved initializing the sharedCSIManager within the scheduler to prevent runtime panics when processing pods with PVCs. This approach mirrored upstream Kubernetes behavior, ensuring consistency and simplifying future maintenance of CSI-related features. By integrating the sharedCSIManager initialization into the scheduler runtime, the changes improved scheduling reliability for PVC workloads and reduced disruptions. The contribution enhanced the codebase’s alignment with Kubernetes CSI handling, supporting more robust and maintainable scheduling for stateful workloads.