Agent-guided workflows to accelerate model customization in Amazon SageMaker AI

This article announces agent-guided workflows in Amazon SageMaker AI that streamline model customization using natural language prompts and AI coding agents.

• AI agents guide entire customization lifecycle from data prep through deployment
• Nine modular skills encode expertise in fine-tuning, evaluation, and deployment
• Supports three techniques: SFT, DPO, and RLVR with automatic recommendations
• Kiro and Claude Code agents available in SageMaker AI JupyterLab
• Generates ready-to-run notebooks for each workflow step
• Skills customizable to match organizational standards and governance
• Reduces months of specialized ML work to days
• Supports deployment to SageMaker endpoints or Amazon Bedrock

SageMaker AI agent skills democratize model customization by automating complex workflows, enabling faster time-to-production for domain-specific models without deep ML expertise.