We will now put together the various templates required to deploy and train our model at scale. These templates include:

• AWS cloud formation templates: Virtual instances and related resources
• Kubernetes or KOPS configuration: K8s cluster management
• Docker templates or Makefile: Create images to deploy on our K8s cluster

We are choosing a particular path here. AWS has services such as Elastic Container Service (ECS) and Elastic Kubernetes Service (EKS) that are accessible via simple API calls. Our purpose here is to engage with the nitty-gritty details, so that you can make informed choices about how to scale the deployment of your own use case. For now, you have greater control over container options and how processing is distributed, as well as how your model is called when deploying containers to a vanilla EC2 instance. These services are also expensive, as we'll see in a later section regarding cost and performance trade-offs when making these decisions.