Amazon SageMaker Autopilot
aws.amazon.com
- Free plan
- Web
- API
What Amazon SageMaker Autopilot is
Amazon SageMaker Autopilot automatically builds, trains, and tunes machine learning models from data, with visibility into the process. It is now available in SageMaker Canvas, where models can also be deployed to production with one click. Autopilot infers whether data suits binary classification, multiclass classification, regression, or time series forecasting, then presents generated models in a leaderboard with performance measures such as accuracy and precision. It can fill missing values, provide statistical insights about dataset columns, and extract information from non-numeric columns. Users can combine their own transformations with more than 300 preconfigured options, and customize data splits and training settings. Autopilot can generate a SageMaker Studio Notebook for inspection, refinement, and recreation of a model. SageMaker Studio lists it alongside services including Pipelines, Experiments, Debugger, Model Monitor, Clarify, and JumpStart. SageMaker AI charges by usage, and on-demand pricing has no minimum fees or upfront commitments.
Who it suits
Autopilot suits people building predictive models who want automated model generation alongside control over data preparation and training options. It is relevant for use cases such as price or churn prediction, risk assessment, and forecasting.
Good to know, and what to watch
- Builds, trains, and tunes models automatically
- Compares models in a metrics leaderboard
- Supports four prediction problem types
- Offers more than 300 preconfigured transformations
- Can generate an inspectable SageMaker Studio Notebook
- No free plan is listed
- Free Tier Canvas hours last only two months
Plans
Every plan and priceOn-demand
Savings Plans
On-demand SageMaker AI
Amazon SageMaker AI pay-as-you-go
From $1.90
Getting started
- Visit https://aws.amazon.com/sagemaker/ai/autopilot/.
- Use the web experience or API.
- Start with the free trial, or use SageMaker AI pay-as-you-go.
Limits to check
The SageMaker AI Free Tier includes 160 hours per month of Canvas session time for the first two months after the first month in which a customer creates a SageMaker AI resource. The listed plan charges for usage; no Autopilot-specific rate is stated.
The Club's review
Amazon SageMaker Autopilot builds and tunes machine-learning models from your data, with options to inspect and adjust the process. It suits teams that want automated model development within SageMaker Canvas; buyers should weigh its usage-based charges and cloud deployment against their need for flexibility and cost control.
Overview
Autopilot is part of SageMaker Canvas and helps move from prepared data to candidate models and production deployment. It is strongest when automation is a starting point rather than a black box: generated notebooks, model comparisons, and adjustable training choices give practitioners ways to review the work. Its place in the broader SageMaker ecosystem is useful for AWS-centered teams, but it is not a standalone local tool.
Key features
Automated modeling and comparison
Autopilot infers whether a dataset suits binary classification, multiclass classification, regression, or time series forecasting, then builds, trains, and tunes models. A leaderboard ranks the resulting models and presents metrics such as accuracy and precision. That makes comparison more transparent than accepting one automatic result, though users still need to decide which metrics suit their objective.
Preparation and control
Preparation features can fill missing data, provide statistical insights about columns, and extract information from non-numeric data. Users can combine their own transformations with more than 300 preconfigured transformations, and adjust data splits and training options. This supports teams that need to shape the workflow, while giving less experienced users a substantial set of prepared steps to start from.
Inspection and deployment
Autopilot can generate a SageMaker Studio Notebook for a model, which users can inspect, refine, and use to recreate it. Models can be deployed to production with one click through SageMaker Canvas. These options make it easier to move beyond an automatically generated result, although deployment remains within AWS's cloud workflow.
In SageMaker Studio, Autopilot sits alongside Pipelines, Experiments, Debugger, Model Monitor, Clarify, and JumpStart. AWS also provides support plans with proactive guidance, issue resolution, and tools. Its cloud security approach combines security of the infrastructure with customer controls in the cloud, and AWS provides compliance resources for customers.
Pricing
Amazon SageMaker AI pay-as-you-go: free. Charges are based on usage; on-demand pricing has no minimum fees or upfront commitments. Buyers should account for the usage basis rather than assume a fixed subscription cost.
SageMaker AI has no free plan, but its Free Tier covers the first two months from the first month a customer creates a SageMaker AI resource. It includes 160 hours per month of Canvas session time during that period. This is a time-limited way to evaluate Canvas, not an ongoing free allowance; teams expecting regular use should plan for paid usage after the initial window. A free trial is available.
Platforms
Autopilot is available through web and API interfaces, with cloud deployment and vendor-cloud hosting. Its workflow interface is both visual and API-oriented, giving teams a visual route through Canvas as well as programmatic access. It also supports feature engineering, automated model selection, model explainability, data profiling, visual workflows, scheduled runs, streaming ingestion, SQL analytics, governance catalogs, online and offline stores, point-in-time joins, and open table formats. These capabilities make it relevant to broader data and machine-learning workflows, not just model generation.
Who it's for
Autopilot is a good fit for AWS teams that want to automate model building while retaining access to comparisons, transformations, training settings, and notebooks. Its listed use cases include price prediction, churn prediction, risk assessment, and time series forecasting. It is less suitable for organizations seeking a permanently free plan, a fixed subscription budget, or self-hosted deployment.
Pros and cons
- Pros: Automatic task inference and model generation can reduce the work of starting a modeling project.
- Pros: A ranked leaderboard, performance metrics, and generated notebooks leave room to evaluate and refine model choices.
- Pros: More than 300 preconfigured transformations can be combined with user-defined steps, data splits, and training options.
- Cons: Usage-based pricing and an unstated Autopilot-specific rate make costs harder to predict before workloads are known.
- Cons: The Canvas Free Tier lasts only for the first two months and has a 160-hours-per-month session allowance.
- Cons: Cloud deployment makes it a poor match for teams that require self-hosting.
Alternatives
If the priority is a dedicated feature-store product rather than automated model creation, compare Feature Store Software options. For labeling workloads, see AI Data Labeling Tools, AI Image Annotation Tools, or Data Labeling Software. Buyers focused specifically on automated model building can compare AutoML Software; teams seeking preparation tools can also review Data Preparation Software.
Canal is a free, Apache-2.0 open-source option available through API, Linux, macOS, and self-hosted platforms. Feast is a free open-source feature store with API, Linux, self-hosted, and web platforms. Hopsworks Feature Store has a free plan with one project, Feature Store and Model Registry, and community support, while its Enterprise plan uses custom pricing.
OpenMLDB is a free open-source machine-learning database with standalone and cluster versions. JFrog ML Feature Store is a paid option with a free trial and custom pricing. Databricks Feature Store charges monthly based on usage with per-second granularity; rates vary by product, cloud provider, and region. Red Hat OpenShift AI Feature Store requires Red Hat OpenShift and offers self-managed subscriptions with Standard or Premium support. Chronon is a free Apache 2.0 open-source option with no usage limits stated.
Verdict
Choose SageMaker Autopilot if your team works in AWS and wants model-building automation without surrendering visibility into comparisons, transformations, and training choices. Its notebooks and adjustable workflow are the main reasons to prefer it over a black-box approach. Look elsewhere if you need self-hosting, a lasting free plan, or clearer costs before usage begins.
The verdict
SageMaker Autopilot automates model creation while leaving room to inspect notebooks and adjust transformations or training settings. Usage-based charges and the limited-duration Canvas Free Tier are key considerations before adopting it.
Questions people ask
Does Amazon SageMaker Autopilot have a free plan?
No free plan is listed. SageMaker AI’s Free Tier includes 160 hours per month of Canvas session time for the first two months from the first month a customer creates a resource.
Which prediction types does Autopilot support?
It infers whether data suits binary or multiclass classification, regression, or time series forecasting.
Can I inspect or customize generated models?
Autopilot supports custom transformations, data splits, and training options. It can also generate a SageMaker Studio Notebook to inspect, refine, and recreate a model.
Where can I use it?
The listed platforms are API and web. Models can be deployed to production through SageMaker Canvas.
What the maker publishes
- API
- The public Autopilot REST API includes CreateAutoMLJob, CreateAutoMLJobV2, DescribeAutoMLJob, DescribeAutoMLJobV2 and ListAutoMLJobs operations. source
- Automatic model selection
- Autopilot infers problem types such as binary classification, multiclass classification, regression and time-series forecasting, then evaluates many candidate algorithms. source
- Canvas migration
- Since November 30, 2023, the Autopilot user interface has been migrating into SageMaker Canvas, while Studio Classic users can continue using the older interface. source
- Compliance
- AWS provides compliance information and resources for customers to understand AWS services’ compliance status. source
- Customization
- Users can apply their own transformations alongside more than 300 preconfigured data transformations and customize data splits and training options. source
- Data preparation
- It can fill missing data, provide statistical insights about dataset columns, and extract information from non-numeric columns. source
- Data preprocessing
- Autopilot handles missing values, normalizes data, selects features and prepares training data automatically. source
- Deployment
- Models can be deployed to production with one click through SageMaker Canvas. source
- Deployment
- After training, users can deploy Autopilot models to real-time endpoints or run batch inference jobs. source
- Documentation depth
- AWS publishes separate user guides, API references, deployment instructions, algorithm-support pages and quota documentation for Autopilot. source
- Explainability
- Autopilot generates feature-importance and model-performance reports, including confusion matrices and ROC or precision-recall visualizations where supported. source
- Feature engineering
- The product offers more than 300 preconfigured data transformations and allows custom preprocessing and feature-engineering steps. source
- Founded
- 2006 source
- Founded
- 2024 source
- Free tier
- SageMaker AI’s Free Tier runs for the first two months from the first month a customer creates a SageMaker AI resource and includes 160 hours/month of Canvas session time. source
- Generated notebooks
- Autopilot can generate a SageMaker Studio Notebook for a created model so users can inspect, refine, and recreate it. source
- Headquarters
- Seattle, Washington, United States source
- Headquarters
- Seattle, Washington, United States source
- Input quota
- The default input-dataset quota is 100 GB and can be increased to hundreds of gigabytes; a single Parquet file is limited to 2 GB. source
- Integrations
- SageMaker Studio lists Autopilot alongside SageMaker Pipelines, Experiments, Debugger, Model Monitor, Clarify, and JumpStart. source
- Integrations
- Reviewer evidence describes integrations with Amazon S3, IAM, CloudWatch and ECR, while AWS documents Studio and Canvas integration. source
- Kubeflow integration
- AWS provides open-source SageMaker components for Kubeflow Pipelines to use SageMaker for labeling, training and inference steps. source
- Leaderboard
- Experiments produce a ranked model leaderboard with metrics such as accuracy and precision for choosing a deployment candidate. source
- Limit: complexity and interface
- G2 reviewers report that SageMaker can feel overcomplicated, with a cluttered interface and pipeline tooling that is less friendly than alternatives. source
Checked against the maker's pages on 19 Oct 2026.