Worked on the aws/sagemaker-hyperpod-cli repository to deliver a feature upgrade for the Health Monitoring Agent, focusing on improving GPU monitoring reliability. Addressed the issue of false positive GPU alerts by introducing a two-minute delay before initiating nvml component checks, which reduced unnecessary alerts and manual triage. Updated Helm chart image tags to align with the new agent version, ensuring consistency across deployments. Utilized DevOps practices and Kubernetes orchestration, with YAML as the primary configuration language. The work enhanced the reliability of GPU health monitoring, streamlined release management, and contributed to more stable and predictable monitoring in production environments.