Nikhil Deshpande developed a feature for the NVIDIA/gpu-operator repository that enables dcgm-exporter metrics to be exposed via the host network, allowing external Prometheus instances to scrape GPU metrics from pods outside the Kubernetes overlay network. He implemented this by configuring the dcgm-exporter daemonset to use hostNetwork, bridging the gap between internal cluster monitoring and external observability tools. Using Go and leveraging his DevOps and Kubernetes expertise, Nikhil addressed the challenge of monitoring GPU workloads across both on-premises and cloud environments. This work improved metrics visibility and incident response readiness, demonstrating a focused and technically sound approach within a short timeframe.