Plaid has successfully migrated its production services from Elastic Container Service (ECS) to Kubernetes. The company initially used ECS but faced operational issues such as complexity in creating new services and lack of visibility into service performance. After a two-year migration process, Plaid implemented a self-serve deployment model called Self-Serve or Scaffold, which automates much of the deployment process and allows teams to operate their own services. However, the company encountered scaling challenges with many services requiring more pods than could fit in a single cluster. To address these issues, Plaid is now redesigning its Kubernetes clusters for multicluster deployments, allowing traffic to shift between clusters if one fails or needs maintenance.