Auto-Sklearn is an open-source automated machine learning (AutoML) toolkit built on scikit-learn, published on GitHub with around 8,100 stars. Developed by the University of Freiburg’s AutoML group, it automatically finds a strong ML pipeline for your dataset — model, preprocessing, and hyperparameters included — within a time budget you set.
Auto-Sklearn searches a large configuration space of scikit-learn estimators and preprocessing steps using Bayesian optimization powered by SMAC. Two ideas make it much stronger than naive search. First, meta-learning: it computes meta-features of your dataset and warm-starts the search with configurations that worked well on similar datasets from a large offline benchmark. Second, after the search it automatically builds an ensemble from the best pipelines found, which reliably beats any single model. The user experience is deliberately simple — you call fit() with a time limit and a per-run limit, and Auto-Sklearn handles resampling strategies, cross-validation, and pipeline selection in the background. It focuses on classification and regression tasks on tabular data and remains a reference implementation in AutoML research, actively maintained with updates into 2026.
Auto-Sklearn is free and open source under the BSD-3-Clause license. There is no paid version or hosted service; you run it on your own machines.
Auto-Sklearn is for data scientists who need a strong tabular ML baseline fast, Kaggle-style competitors prototyping solutions, and teams that want automated pipeline search without adopting a proprietary AutoML platform.