Editorial illustration for Amazon adds agentic fine‑tuning to SageMaker for Llama, Qwen, Deepseek, Nova
Amazon adds agentic fine‑tuning to SageMaker for Llama,...
Fine-tuning an AI model has been grueling, specialist work. Now it's over. Amazon declared it so Wednesday with a new agent for SageMaker, its cloud platform for machine learning.
Developers simply state what they need in plain English. The agent handles the rest: selecting the training method, preparing the data, running the job, delivering production-ready code. It launches with immediate support for four major model families—Meta's Llama, Alibaba's Qwen, Deepseek, and Amazon's own Nova—turning a weeks-long engineering slog into a straightforward instruction.
Amazon SageMaker AI now includes an AI agent designed to help developers customize language models.
That preinstalled agent is named Kiro. But crucially, as THE DECODER notes, you can swap it for Anthropic's Claude Code or another. Amazon's goal is removing friction, not enforcing a lock-in.
Bundling this AI mechanic into SageMaker is a shrewd infrastructure play. They make their platform indispensable by making it the easiest. The fundamental shift is stark: from engineering to instruction.
You're no longer a mechanic under the hood. You're a boss giving an order. This will fuel rapid experimentation, yes.
It also creates a dangerous new abstraction, hiding complex choices that—when wrong—will spawn a whole new class of elusive, expensive problems.
Related Reading
Browse all Meta news →
LLMs & Generative AI ChatGPT's 'Nerdy' tweak rewards goblin metaphors in answers, study finds
LLMs & Generative AI Google tests visual 'magazine-style' UI for Gemini 3 Pro users
LLMs & Generative AI AI Engineers Face Rising Costs, Need New Strategies for Efficiency
Related Topic WhatsApp launches Meta AI Incognito Chat, cuts latency for privacy
AI Tools & Apps
Related Topic Nvidia and Meta ink deal; Nvidia touts hardware for inference and AI training
Business & Startups
From Archives LLaMA-Factory Enables UI-Based Fine-Tuning and Multi-Model Support Locally
From Archives Self-Healing Layer Scores Answers in Real Time to Counter RAG Hallucinations
From Archives Zuckerberg commits USD 500 million to AI-driven biology research at Meta
From Archives LlamaIndex CEO: AI scaffolding collapses as models surpass humans on massive data
Common Questions Answered
What is agentic fine-tuning in Amazon SageMaker and how does it simplify model customization?
Agentic fine-tuning is a new feature in Amazon SageMaker that allows developers to fine-tune AI models by simply stating their requirements in plain English, rather than manually handling complex technical processes. The agent automatically selects the appropriate training method, prepares the data, runs the job, and delivers production-ready code, eliminating the need for specialized fine-tuning expertise.
Which model families are currently supported by Amazon's agentic fine-tuning feature?
Amazon's agentic fine-tuning launches with immediate support for four major model families: Meta's Llama, Alibaba's Qwen, Deepseek, and Amazon's own Nova. These diverse model options provide developers with flexibility in choosing the foundation model that best fits their specific use case requirements.
What is Kiro and can developers use alternative AI agents with SageMaker's fine-tuning feature?
Kiro is the preinstalled agent that powers SageMaker's agentic fine-tuning capabilities by default. However, developers can swap Kiro for alternative agents such as Anthropic's Claude Code or others, giving them flexibility in choosing which AI agent manages their fine-tuning workflow while avoiding vendor lock-in.
How does Amazon's agentic fine-tuning approach represent a shift from traditional machine learning workflows?
The agentic fine-tuning approach fundamentally shifts the role of developers from being technical mechanics handling complex engineering tasks to being decision-makers who simply issue instructions in plain English. This transformation removes friction from the model customization process and enables rapid experimentation by making fine-tuning accessible to developers without deep machine learning expertise.
Further Reading
- Agent-guided workflows to accelerate model customization in Amazon SageMaker AI — AWS Machine Learning Blog
- Accelerate agentic tool calling with serverless model customization in Amazon SageMaker AI — AWS Machine Learning Blog
- Amazon SageMaker AI now supports serverless reinforcement fine-tuning for 12 additional open-weight models — AWS What's New
- Advanced fine-tuning techniques for multi-agent orchestration patterns from Amazon at scale — AWS Machine Learning Blog
