Amazon SageMaker AI launches multi-turn reinforcement learning for AI agent model customization

Summary

• Introduces serverless multi‑turn reinforcement learning for fine‑tuning agent models on SageMaker AI
• Provides end‑to‑end training loop, trajectory tracking, and evaluation with integration to Bedrock AgentCore, EKS, EC2, Fargate, etc.
• Available via SageMaker Studio and Python SDK; pay‑as‑you‑go token pricing and supports specific models in US regions