FLAML

Why we included this project

FLAML is a lightweight Python library built to take the manual work out of model selection and hyperparameter tuning. Point its AutoML engine at a tabular dataset with a classification or regression task and it returns a working scikit-learn-style estimator, choosing the model and search budget under whatever time or resource limit you set. The same library exposes a generic tuning API for optimizing parameters of any Python function, which covers everything from XGBoost and LightGBM configurations to inference settings for foundation models, and it copes with large search spaces and uneven evaluation costs. Data scientists get a low-code entry point, while engineers can extend it and plug in MLflow tracking to fit existing MLOps pipelines. A zero-shot mode is a nice touch: it lets LightGBM and XGBoost pick decent hyperparameters straight from your data, so you capture most of the benefit with almost no new code.