Skip to main content
Use this deployment path when you run Poolside on your Amazon Elastic Kubernetes Service (Amazon EKS) infrastructure. You provision the AWS infrastructure (EKS cluster, database, object storage, container registry, and supporting services), then install the Poolside platform with Helm. You manage the platform independently, with Poolside support for installation and configuration.

​
How to use this section

The fastest path to a working deployment is to follow the reference architecture. It is an opinionated, validated configuration for a minimal, production-ready Poolside deployment, and ships a Terraform starting point that provisions the AWS infrastructure, the Kubernetes prerequisites, and the Helm install in a single terraform apply. If you assemble the AWS foundation with your own infrastructure-as-code, the install guide describes the chart’s expectations so you can match them.
  1. Reference architecture: Recommended starting point.
  2. Install the platform on Amazon EKS: Chart-level install steps for self-assembled deployments.
  3. Configure model inference on Amazon EKS: Local GPU inference or external OpenAI-compatible APIs.
For supporting topics, see AWS cost modeling and Deployment security. If you already run the Terraform-based AWS deployment bundle, see Amazon EKS (Terraform) — legacy and the migration guide.

​
Model inference

You can serve Poolside models from GPU-backed inference workloads inside your Amazon EKS cluster. Deploy the inference-stack Helm chart against your own GPU node group and NVIDIA GPU Operator setup, or use the reference architecture, which provisions both for you. You can also connect Poolside to any external OpenAI-compatible model API instead. See Model inference on Amazon EKS for details.