cd /news/artificial-intelligence/announcing-ai-data-preparation-capab… · home › topics › artificial-intelligence › article
[ARTICLE · art-171305] src=aws.amazon.com ↗ pub= topic=artificial-intelligence verified=true sentiment=↑ positive

Announcing AI data preparation capabilities in Amazon SageMaker: Build high-quality datasets for model development workflows

Amazon Web Services expanded Amazon SageMaker AI with new AI data preparation capabilities, delivered through Python SDK APIs and an AWS AI-ML agent skill, so teams can generate synthetic examples, assess and improve datasets, and route low-confidence rows to human review. The capabilities include auto-labeling with a multi-model jury whose agreement becomes each row's confidence score, synthetic data generation, data quality evaluation for duplicates and answer leakage, and data transformation for target training techniques. Human review runs in Amazon SageMaker Studio on the open-source Label Studio Community Edition, with reviewer corrections feeding back into auto-labeling.

by read7 min views6 publishedNov 17, 2026
Announcing AI data preparation capabilities in Amazon SageMaker: Build high-quality datasets for model development workflows
Image: AWS ML Blog

Artificial Intelligence #

Teams that ship generative AI to production know that a model is only as good as its data. That is why preparing high-quality training and evaluation data comes first. Fine-tuning an open-weight model, running reinforcement learning, or building an evaluation suite all depend on data that is accurate, well-labeled, and representative of your task. For narrow, well-defined tasks, a smaller model customized on clean, task-specific data can often match or exceed the accuracy of a larger general-purpose model prompted into shape, at a lower cost per inference.

In practice, though, preparing that data is often the slowest part of the project. Labeled examples are scarce or expensive, and privacy rules limit what you can use. Quality problems such as duplicates, label noise, and answer leakage go unnoticed until a training run disappoints, and every training technique expects data in its own format. As a result, data scientists spend weeks wrangling data instead of building models.

Today, we’re expanding Amazon SageMaker AI with new AI data preparation capabilities for model development workflows. You can generate synthetic examples, assess and improve existing datasets, and bring expert feedback into the process through Python SDK APIs and the AWS AI-ML agent skill. You work in your notebook or IDE, with specifications and code that you can inspect and change.

Prepare data without hand-building pipelines #

With AI Data Preparation capabilities, you don’t need to stitch together labeling tools, generation scripts, and validation jobs. Each capability is a standalone building block. You can use one of them on its own or chain them together according to the workflow you want.

  • Auto-labeling: Label data at scale from an unlabeled dataset, a small labeled seed set, or no labels at all. Built-in task types include classification, named entity recognition, extractive question answering, ordinal rating, and summarization, and you can define custom tasks. You describe the task in a short labeling spec, which is validated against your own examples before the full dataset is labeled. For tasks with a fixed set of possible answers, such as classification, named entity recognition, and extractive question answering, a jury of several models labels each row, and their level of agreement becomes that row’s confidence score.
  • Human review: Rows below your confidence threshold go to a review queue in Amazon SageMaker Studio, which uses the open-source Label Studio Community Edition. Outside subject matter experts can access the review app through the link in the partner app box in SageMaker Studio. Reviewer corrections feed back into auto-labeling, and you can generate consensus metrics and labeling reports.
  • Synthetic data generation: Generate training, evaluation, and test data from a specification. You can ground the output in your own seed data and documents, and the service generates a code-based data generation pipeline that you can review, modify, reuse, and scale.
  • Data quality evaluation: Check datasets against your own business rules, built-in metrics, and LLM-as-a-judge metrics. Find duplicates and near-duplicates, and surface issues you didn’t think to check for, such as label noise or answer leakage.
  • Data transformation: Clean and reshape data for your target training technique. Describe the change in plain language and have the code generated for you, use built-in operations such as dedupe, merge, split, and filter, or run your own Python at scale.

You work wherever you already work. Use the SageMaker Python SDK directly or provide your coding agent the AI-ML skill in the agent toolkit and let your agent invoke the APIs according to established data science best practices. Describe what you need, and the agent selects and runs the right capability in SageMaker Studio JupyterLab, Kiro, or your local IDE with your coding assistant of choice.

For teams building specialized models, the hardest part is often finding enough task-specific data. Splunk, a Cisco company, faced exactly this when building evaluation models for its Agent Observability product. “For Splunk Agent Observability, our ability to train robust GenAI Eval models relies on access to high-quality, diverse datasets which are hard to curate for very specific tasks. Amazon SageMaker AI data preparation capabilities for synthetic data generation is meaningfully helping us scale these capabilities. We’ve been able to generate statistically representative synthetic datasets reliably and fast, while preserving the underlying patterns our models depend on. The seamless integration with the rest of our SageMaker workflow means our data scientists and customers spend less time wrangling data and more time building models that drive real enterprise value through Splunk.”

— Yash Sheth, Director of Engineering, Splunk Agent Observability

Chain capabilities from raw data to training #

The value grows when you combine the capabilities. Some common patterns:

Build a golden evaluation dataset: Auto-label your raw data from a small seed set. High-confidence rows are accepted automatically, and low-confidence rows go to human reviewers. Their corrections improve the next labeling pass, resulting in a trusted, human-validated dataset to benchmark your models against.

Go from raw documents to a trained model in one workflow:

  1. Generate data: Synthetic data generation creates task-specific examples grounded in your documents and seed data.
  2. Transform it: Data transformation reshapes the data into the format your training technique expects, such as supervised fine-tuning (SFT), direct preference optimization (DPO), or reinforcement learning (RL).
  3. Validate it: Data quality evaluation checks the dataset for duplicates, leakage, and rule violations before you spend compute on training.
  4. Train: Train with SageMaker AI model customization.

How it works #

Consider a team preparing data to evaluate agentic customer support assistants. Its product guide says that a factory reset erases saved settings. The team wants to test whether the assistant preserves that warning when answering reset questions.

One pass through the preparation workflow looks like this:

  1. Define the requirements: Supply the product guide as context and describe the questions and reference answers you need. Include the requirement that answers about a factory reset preserve the warning about saved settings. Review the specification produced with your coding agent.
  2. Generate and inspect a sample: Generate a small batch and check the records against the guide. Review the generation and check the code. Confirm that the examples cover the intended questions and that the reference answers follow the source material.
  3. Assess and review: Run a data quality assessment and inspect the record-level findings. Ask a subject matter expert to review ambiguous answers or disputed assessments. Use the results and annotations to request improvement recommendations.
  4. Apply changes and repeat: Revise the specification if the requirements need clarification or transform the dataset to address issues such as duplicates. Generate additional examples where needed, then reassess the revised dataset before using it to evaluate the assistant.

This example workflow illustrates how the capabilities fit together.

Keep control of your preparation workflow and data #

With the AI-ML skill, you describe the task in natural language while your coding agent helps build specifications, prepare code, and coordinate the data preparation steps. You can open those artifacts to see how the workflow is configured and make changes directly.

If you prefer to invoke the APIs yourself, use the Python SDK APIs to call individual capabilities from your own code. The SDK APIs take a specification as input, so we recommend creating it with the help of an agent. You can add quality assessment to an existing process, generate data for a new task, or combine several operations. You choose the preparation steps your project needs and how they fit into your model customization and evaluation processes. AI Data Preparation runs with an AWS Identity and Access Management (IAM) execution role that you control. Your source data, generated datasets, and labels stay in your own Amazon Simple Storage Service (Amazon S3) buckets, in the AWS Region you choose.

Get started today #

The AI data preparation capabilities are generally available in the US East (N. Virginia) and US West (Oregon) Regions.

There are two ways to get started:

  • With your coding agent – Set up thedata preparation skills in your supported notebook or IDE. Describe one training or evaluation task and review the resulting specification before generating a sample.
  • With Python – Use the SageMaker Python SDK directly to run individual data preparation capabilities from your own code. The SDK takes a specification as input, so we recommend creating your specifications with the AI-ML skill first, then using the SDK to run them in your workflow.

For more information, see the Amazon SageMaker AI documentation.

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @amazon web services 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/announcing-ai-data-p…] indexed:0 read:7min 2026-11-17 · —