We're doing most of our ML in python right now. We're probably heading towards a unified scala+spark platform in the next year or two though.

Apache Spark is what I use, personally.