Welcome to your second lesson in managing ML resources with the SageMaker AI Console. In the previous lesson, you learned to navigate and explore your console environment—finding training jobs, checking endpoints, and understanding what's currently running. Now you'll learn to actively manage these resources by making updates, removing unused components, and keeping your environment clean and cost-effective.

Our focus is on four essential management operations: updating endpoint configurations to optimize performance and costs, deleting unused endpoints, removing endpoint configurations that are no longer needed, and cleaning up obsolete models from your registry. By the end, you'll be able to confidently modify running deployments and maintain a well-organized ML infrastructure.

One of the most common management tasks you'll perform is updating endpoint configurations to better match your actual needs. Whether you need more computational power for increased traffic or want to reduce costs by right-sizing over-provisioned resources, the SageMaker console makes these updates straightforward.

The following video demonstrates how to access an endpoint's configuration and make runtime updates through the console interface.

Understanding how to update endpoint configurations gives you the flexibility to adapt your deployments as requirements change, ensuring optimal performance while controlling costs.

When endpoints are no longer needed, removing them is one of the most effective ways to reduce AWS costs immediately. Once deleted, the endpoint stops immediately and can't be recovered, though you can recreate it later using the same model and configuration.

Watch the next video to see how to identify and delete unused endpoints through the console safely.

Mastering endpoint deletion helps you maintain cost-effective deployments by removing resources that are no longer serving a business purpose.

After you delete an endpoint, the endpoint configuration that defined its deployment settings still exists in your environment. These configurations specify instance types, scaling settings, and other deployment parameters, but once they're no longer used by any endpoints, they become clutter that should be removed to maintain organization.

Before deleting an endpoint configuration, you need to ensure no active endpoints are using it. The console shows you which endpoints, if any, are associated with each configuration. Once you've confirmed a configuration is unused, deletion is permanent and immediate.

The following video demonstrates how to identify unused endpoint configurations and remove them safely.

Keeping your endpoint configurations clean makes it easier to find and reuse the configurations you actually need for future deployments.

After you delete endpoints and their configurations, the underlying models still exist and consume storage space in your model registry. While storage costs are relatively low, a cluttered registry makes it difficult to identify current, approved models and can lead to accidentally deploying outdated versions.

Model deletion requires extra caution because it permanently removes the trained artifacts. Before deleting any model, verify that no endpoints are currently using it and that you won't need it for future comparisons or rollbacks.

Watch the next video to learn how to clean up obsolete models from your registry.

Regular model cleanup ensures your registry remains organized and focused on the models that matter for your current projects.

You now have the essential skills for active ML resource management in SageMaker. You can update endpoint configurations to optimize performance and costs, delete unused endpoints to reduce expenses, remove obsolete endpoint configurations to maintain organization, and clean up old models to keep your registry focused. These management capabilities build on the navigation skills from the previous lesson to give you more control over your ML infrastructure.

In the upcoming practice session, you'll apply these management skills to realistic scenarios. You'll update endpoint configurations to handle changing requirements, identify and remove unused resources to reduce costs, and clean up obsolete components to maintain an organized environment. This hands-on experience will build your confidence in managing real SageMaker deployments and prepare you for the resource management challenges you'll face in production ML projects.