Nearly every machine learning estimator has parameters assumed to be given or "hyper-parameters". Finding the optimal set of hyper-parameters is a difficult and time-consuming process for most modern estimators. A recent breakthrough algorithm, Hyperband addresses this problem and has strong theoretical backing. Hyperband quickly finds high scores by intelligently choosing which hyper-parameters to evaluate. This algorithm is now implemented in Dask-ML, and can be applied to many estimators. This talk will explain what Hyperband does and the features it provides, and walk through an example that guided development.