Using it to run easily-scalable Spark clusters. Previously used it for large distributed builds. It's been pretty great (even if annoying at times).