Amazon SageMaker AI launches AI agent experience for model customization

Amazon Web Services has launched a new AI agent experience within SageMaker AI that dramatically reduces model customization time from months to days or hours. The agentic experience leverages natural language interactions with coding agents to streamline the entire machine learning workflow, from use case definition to production deployment. Developers can now work through complex processes like data preparation, model selection, fine-tuning experiments, and deployment configuration using conversational interfaces rather than manual infrastructure management. The new capability is built on SageMaker AI model customization agent skills that provide expertise in fine-tuning, data format transformation, quality evaluation using LLM-as-a-judge metrics, and deployment options to Amazon Bedrock or SageMaker endpoints. The system integrates with popular IDEs including Visual Studio and Cursor, and supports multiple coding agents such as Kiro, Claude Code, and GitHub Copilot. It can optimize major model families including Amazon Nova, Llama, Qwen, and GPT-OSS while generating reusable, editable code artifacts for transparency and integration into AIOps pipelines.