Welcome to your third lesson in managing ML resources with the SageMaker AI Console. You've learned to navigate and explore resources, then actively manage existing deployments. Now you'll learn the basic of the deployment workflow—turning your completed training jobs into live endpoints that serve predictions.

Our focus is on three core deployment operations: creating models from training jobs, building endpoint configurations that define deployment settings, and launching endpoints that serve your models in production. Let's dive in!

Before you can deploy a model to serve predictions, you need to create a model resource that packages your trained artifacts with the appropriate inference container. This model becomes the deployable unit that SageMaker can load onto compute instances.

One easy way to create a model is from a completed training job that has already produced model artifacts stored in S3. You'll combine these artifacts with an inference container image, creating a complete package ready for deployment.

The following video demonstrates how to create a model from a training job through the console interface.

Creating models properly ensures your trained artifacts are packaged correctly for deployment with the right inference environment.

Endpoint configurations serve as deployment blueprints that define how your model will run in production. These configurations specify compute resources, scaling settings, and traffic routing rules that determine performance and cost characteristics.

You'll choose deployment options based on your traffic patterns and latency requirements. Different configurations provide various performance characteristics while controlling costs based on your specific needs.

Watch the next video to see how to create endpoint configurations that optimize for your specific deployment requirements.

Well-configured endpoint settings ensure optimal performance while controlling costs based on your actual usage patterns.

The final step transforms your model and configuration into a live endpoint serving predictions. This process provisions compute resources, loads your model, and makes it available for inference requests.

The following video demonstrates the complete endpoint creation workflow.

Successful endpoint deployment ensures your model is ready to serve predictions reliably in production environments.

You now have the complete deployment workflow for taking approved models from development to production. You can create models from training jobs, build endpoint configurations that optimize for your requirements, and launch endpoints ready for production use. These deployment skills complete your SageMaker resource management capabilities.

In the upcoming practice session, you'll deploy models end-to-end using realistic scenarios. This hands-on experience will build your confidence in production ML deployments and prepare you for real-world deployment challenges.