Transform AI development with new Amazon SageMaker AI model customization and large-scale training capabilities

This article announces major Amazon SageMaker AI advancements for model customization and large-scale training capabilities.

• Serverless model customization with AI agent-guided workflow using natural language
• Supports supervised fine-tuning, DPO, RLAIF, and reinforcement learning techniques
• Amazon Nova Forge enables building frontier models from early checkpoints with proprietary data
• Elastic training automatically scales based on compute resource availability
• Checkpointless training recovers from failures in seconds, achieving 95% training goodput
• Serverless MLflow provides experiment tracking without infrastructure management
• Features available worldwide in AWS Regions today

SageMaker AI now provides end-to-end capabilities spanning model customization, large-scale training, and production deployment, accelerating AI development from months to days.