Dask itself (and its commercial entity Coiled) seems to be investing heavily in the PyData/ml library ecosystem. I feel like it's about time until they reach feature parity, which would be a huge win for existing users of the PyData stack.