If you need standalone production job, you can use...

# random

[khud]

If you need standalone production job, you can use Java Spark API almost without any problems. But, I think some wrappers are better choice. If you want to do some ad hoc computations you need to have a REPL.