Amazon announced the introduction of AI agent capabilities in its machine learning platform, Amazon SageMaker, aiming to lower the barrier for developers to customize language models and restructure the model development process. As one of Amazon's core AI infrastructures, this upgrade allows developers to trigger an end-to-end modeling process by simply describing the use case in natural language, without manually handling complex API calls and data format conversions.
The AI agent can automatically complete key stages of model development, including recommending training strategies, data preparation, training task scheduling, and result delivery, ultimately outputting complete code in the form of a Jupyter Notebook, which supports subsequent editing and reuse. At the execution level, the system includes an internal agent tool called Kiro AI and provides nine preset "skills," covering the entire lifecycle from dataset inspection to model deployment. Meanwhile, developers can also integrate third-party agents such as Claude Code to adapt to different development habits.
In terms of model compatibility, the agent supports multiple mainstream open-source and commercial model series, including Llama, Qwen, DeepSeek, and Amazon's own Nova, demonstrating the platform's open strategy toward a multi-model ecosystem.
Overall, the integration of AI agents into SageMaker marks the evolution of machine learning development from "toolchain-driven" to "agent-driven." By automating orchestration and using natural language interaction, it significantly shortens the model development cycle while strengthening the cloud platform's central position in the AI productivity toolchain.
