sagemaker-ai Claude Code Plugin
Claude Code PluginClaude CodeDevelopmentAI Agents & AI App DevelopmentCloud, Deployment & CI/CDThe explanation below is AI-generated. Please verify it against the sources.
sagemaker-ai is a plugin that brings AWS machine-learning expertise for Amazon SageMaker AI into coding assistants such as Claude Code. According to its README, it offers a set of skills that walk a user through fine-tuning foundation models, covering use-case definition, dataset preparation, training configuration, evaluation, and endpoint deployment. It also bundles operational skills for Amazon SageMaker HyperPod training clusters, letting a user run remote commands, compare software versions, and diagnose cluster, node, network, and scheduler problems. A companion MCP server named aws-mcp is included to look up AWS documentation and standard operating procedures. Using the plugin requires configured AWS credentials with permissions to call SageMaker and related AWS APIs.
About the service
Amazon SageMaker AI is AWS's managed service for building, training, and deploying machine learning and foundation models, including HyperPod clusters for large-scale distributed training. It sits within the broader SageMaker Unified Studio experience, which also connects to services such as Amazon Bedrock, Redshift, and S3 for generative AI and analytics workloads.
What you can do with sagemaker-ai
- Build a step-by-step plan for a model customization project (planning skill)
- Organize project directories and artifacts (directory-management skill)
- Define a use case's goals, stakeholders, and success criteria (use-case-specification skill)
- Validate dataset quality and convert data into SageMaker-compatible formats (dataset-evaluation, dataset-transformation skills)
- Choose a fine-tuning technique and base model, configure hyperparameters, and run training jobs (finetuning-setup, finetuning skills)
- Design evaluations, select benchmarks, and compare models, including LLM-as-a-judge (model-evaluation skill)
- Configure and launch a deployment endpoint on SageMaker or Amazon Bedrock (model-deployment skill)
- Run remote commands on HyperPod nodes via SSM, check for software version drift, and generate issue reports
- Diagnose HyperPod cluster-wide, node-level, NCCL, performance, and Slurm scheduler problems
- Retrieve AWS documentation and standard operating procedures through the aws-mcp server