Model Deployment

Takes a SageMaker Serverless Model Customization fine-tuned model and generates the code to actually deploy it, either to a SageMaker endpoint or Bedrock. It's methodical about walking through the decision tree: pulls metadata from your training job, figures out whether you've got Nova or an OSS model, shows you eligible targets, then generates pathway-specific deployment code. The workflow is deliberately slow, one confirmation at a time, which feels right given how easy it is to spin up expensive infrastructure by accident. Only handles LoRA fine-tuned models, not full fine-tuning, and it'll check the license terms before deploying. Honest take: this is the boring bridging work between "my model trained" and "I can call my model," automated.

Install to Claude Code

npx -y skills add awslabs/agent-plugins --skill model-deployment --agent claude-code