Contributed to the NVIDIA/gpu-operator repository by developing a sandbox device plugin that enables GPU Feature Discovery pods to launch with the necessary privileges and environment variables, improving the accuracy and visibility of GPU feature detection within Kubernetes clusters. Addressed operational resilience by implementing a permissions fix that grants update rights to NodeFeatures, ensuring device plugins function correctly during node restarts and minimizing disruption to GPU workloads. Leveraged expertise in Kubernetes, containerization, and YAML configuration to enhance workload scheduling fidelity and reduce downtime during maintenance. The work demonstrates a strong grasp of Kubernetes device plugin architecture and practical DevOps problem-solving.