I wrote this post describing what I think is the simplest stack that supports scalable ML.
I'm curious what people with related interests think.
Thanks,
Stefan