It's quite complete for anything numerical/machine learning/statistical at this point, with almost 2000 packages. This is in addition to the very comprehensive standard library, which includes everything you get in Numpy/Scipy. Of course, outside of numerics, you're pretty much out of luck, which is not the case for Python.

It also has multiple dispatch like Common Lisp, as well as gradual typing. All functions are multimethods, actually. This is an enormous abstraction advantage over Python. It's much easier to do complicated things without pulling your hair out in comparison to Python.

I use it a fair amount in my job as a data scientist. It's also what I reach for if I need to write some custom algorithm myself that needs to be high performance, rather than doing it in C.

Does it have a tooling for deployment? Is there a good way to deploy a machine learning model in production?