Managing 18 Production EKS Clusters

Challenge

Nielsen Media runs a diverse portfolio of applications on Kubernetes, and platform reliability was a bottleneck. A small team needed to operate 18 production EKS clusters across multiple environments without sacrificing deployment speed or stability.

Solution

I built a standardized platform using Terraform for cluster provisioning, Helm charts for add-ons (Prometheus, Grafana, ingress controllers, cluster autoscaler), and Argo CD for GitOps-based application delivery. DORA metrics were instrumented to measure deployment frequency, pipeline stability, and change failure rates.

Key Engineering Decisions

• Modular Terraform so each cluster could be spun up consistently while still allowing per-application overrides.
• GitOps with Argo CD so cluster configuration and application state were version-controlled and auditable.
• Rolling deployments with readiness gates to ensure AI workloads could be updated without dropping requests.

Outcome

The platform supported continuous delivery for AI and microservices workloads across production and non-production environments, with measurable improvements in deployment reliability.