At Sarus Technologies, we use dask to test many machine learning models and tune their hyper-parameters in parallel. It's actually quite simple to deploy and scale a dask cluster, thanks to kubernetes and public clouds. If you are interested, check this tutorial out.