Amazon SageMaker AI introduces new feature for rapid model customization

Amazon SageMaker AI has unveiled a new feature that significantly accelerates the process of customizing AI models, reducing what typically takes months to just days or hours. When developing an AI solution, clients must delineate their goals and criteria for success, prepare data, select appropriate models, and conduct numerous experiments using various models and fine-tuning techniques. Once an ideal model is identified, they must then determine the most cost-effective way to deploy it. Throughout this process, customers often face the challenge of managing the infrastructure required to train and deploy these models.

The newly introduced capability allows developers to engage with coding agents through natural language, simplifying the entire journey from defining a use case to deploying a high-quality model in production. This agent-driven experience leverages SageMaker AI’s model customization skills, offering expertise in fine-tuning for specific use cases, transforming data into necessary formats, conducting thorough quality evaluations using LLM-as-a-judge metrics, and providing flexible deployment options to Amazon Bedrock or SageMaker AI endpoints.

Customers can integrate these skills into any Integrated Development Environment (IDE) of their choice, such as Visual Studio or Cursor. Developers have the option to collaborate with multiple coding agents, including Kiro, Claude Code, and CoPilot, to optimize popular model families like Amazon Nova, Llama, Qwen, and GPT-OSS. The experience also generates reusable and editable code artifacts, ensuring transparency, reproducibility, and automation through integration into AIOps pipelines.

Users can install SageMaker AI skills in their preferred IDE using the sagemaker-ai agent plugin. These model customization skills are also pre-installed in SageMaker Studio Notebooks, along with the Kiro coding agent. To begin, users simply need to subscribe to Kiro, open the chat window in Studio Notebooks, and start interacting with the agent to create the workflow.

The experience supports advanced customization methods, including supervised fine-tuning for instruction tuning, direct preference optimization for adjusting tone and preference selections, and reinforcement learning for scenarios requiring verifiable correctness. For more information on model customization with the AI agent experience in Amazon SageMaker AI, users are encouraged to consult the SageMaker model customization documentation.