The plan is to package python modules in jar files
Then use -libjars to distribute them to hadoop nodes.
Then use sys.path.append [trick](http://www.jython.org/jythonbook/en/1.0/ModulesPackages.html#sys-path) to instantiate them

From there can use some usual java->jython calling [convention](http://www.jython.org/jythonbook/en/1.0/JythonAndJavaIntegration.html) to call map & reduce.