Azure AutoML
If your data already lives in Microsoft's cloud, the shortest route to a model is usually the one built into it. Azure AutoML is the automated model-building service inside Azure Machine Learning: you register a dataset, name the target column and the task, set a budget, and it returns a ranked list of trained models.
What it covers
- Task types — classification, regression, time-series forecasting, plus vision and text tasks that fine-tune pretrained models.
- Featurisation — automatic imputation, encoding and text vectorisation, and for forecasting, lag and rolling-window features with rolling-origin validation.
- Search and ensembling — guided selection over scikit-learn, LightGBM and XGBoost learners, finished with voting and stacked ensembles that usually top the leaderboard.
How you drive it
The same job runs three ways: the no-code wizard in Azure ML Studio, the azure-ai-ml Python SDK, or a YAML job submitted through the CLI. Every trial is logged as an MLflow run, so the winning model arrives with its metrics, feature-importance explanations, and a one-click path to a managed endpoint.
Why it matters
The automation itself is competent rather than exceptional — open-source frameworks score comparably on public benchmarks. What the platform adds is the apparatus around it: role-based access to the data, quota-controlled compute clusters that shut themselves down, audit trails, and deployment. For a team already inside Azure, that integration decides the tool choice far more than leaderboard position does.
Azure AutoML is the automated model-building service in Azure Machine Learning. Given a registered dataset, a target column, a task type and a time budget, it featurises the data, trains and ranks candidate models, and builds voting or stacked ensembles, logging every trial as an MLflow run. It matters because the winning model inherits the platform's access control, managed compute and one-click deployment — what makes automated results usable in a governed environment.
Most large companies already keep their data in a cloud service, and Microsoft's is called Azure. Azure AutoML is the button inside it that says, in effect, build me a model from this table.
You pick the spreadsheet, point at the column you want predicted, and say how long the computer may spend on it. The service tidies the data, tries many different methods, combines the best ones, and shows you a scoreboard of what worked. Because all of this happens inside the company's own cloud account, the permissions, the billing and the final step of putting the model to work are handled in the same place — which is usually why a team picks it.