imho you can use python for everything. Some fields make more sense, others make less sense.
I wouldn't implement a large scale equation solver in python, but rather in C or Fortran.
But you have the opportunity to use every java package with jython or you can port the crucial
parts with respect to performance to C/Fortran. Same works for IronPython and .net.
If you wanna see what is possible, have a look at numpy, scipy, matplotlib or sympy.
Those are more science related packages but you can find the same quality in other fields.
Python is easy but powerfull. Just compare the way Python deals with Reflection with the Java way
for example. Same holds for lists ...