We wrote a framework that forces people to write logic as python functions. This made it really easy to offload, i.e. wrap these functions, to Dask for distributed computation. In particular, if you are using Pandas and use the API that Dask implements, you get seamless scaling without having to change any logic. See our hello world example - hamilton/examples/dask at main · stitchfix/hamilton · GitHub.
@skrawcz Welcome to Discourse, and thanks for sharing!