Ably optimizes AI development environments with AWS Agent Sandbox

Fashion commerce company Ably has implemented Amazon Web Services (AWS) Agent Sandbox to streamline the operation of its internal AI development platform. The company utilizes Amazon Elastic Kubernetes Service (Amazon EKS) to provide isolated environments where employees can develop web applications using natural language inputs.

Managing the lifecycle of these ephemeral environments proved challenging. Ably required a system that could automatically provision resources when a project started and reclaim them once the tasks were completed. After evaluating standard Kubernetes workloads, such as Deployments and Jobs, the company determined that these methods lacked the built-in lifecycle control needed to manage temporary development workspaces effectively.

By adopting Agent Sandbox, Ably has delegated the orchestration of environment creation and termination to specialized controllers. The architecture integrates Karpenter to handle node provisioning and scaling based on current workload demands. This approach allows Ably to maintain strict resource separation for projects while ensuring that idle infrastructure is automatically shut down, significantly improving operational efficiency.