I've been trying to change all of my Luigi pipeline tasks from using Pandas to Dask, so that I can push a lot more data through. Seems like an easy process so far, and I like the easy implementation of parallel computing.