Worked on the NVIDIA/gpu-operator repository to enhance observability and configurability of monitoring for GPU workloads. Developed a new ServiceMonitor configuration structure using Go and Kubernetes, enabling users to define custom labels, intervals, and relabeling options for GPU Operator resources. This approach reduced the need for post-deployment tuning and allowed for more granular filtering and alert routing within Prometheus, improving visibility and incident response. The work focused on DevOps practices, emphasizing maintainability and flexibility in monitoring configurations. No bugs were fixed during this period, with efforts concentrated on delivering this feature to support better SLA reporting and operational insight.