Ray uses it to more efficiently transfer objects across processes and nodes.

https://docs.ray.io/en/master/serialization.html

https://arrow.apache.org/blog/2017/10/15/fast-python-seriali...

Here's some material from the official website: https://arrow.apache.org/powered_by/ https://arrow.apache.org/use_cases/

I've also written a bit about this here, from the point of view of a data scientist: https://www.robinlinacre.com/demystifying_arrow/