Hi Michele,
        
        
        As you've observed, jpy was not a library we were aware of
          when we did the evaluation of java-python solutions. I'd be
          interested in any insights you have with regard to its
          approach. We have settled on Py4J as the java-python bridge
          which we are settling down with, building a house & having
          babies... :-)
        
        
        Here is the criteria we used to evaluate the bridges at the
          time, it might be useful to you should you chose to do your
          own evaluation for specific use-cases.
        
        
        Technical merit:
        --------------------
        What is overall architecture, who manages the interpreter
          lifecycle? (e.g. Py4J has separate Python/Java processes, uses
          sockets)
        Does the implementation provide full support for
          C-libraries or conditional/partial support? (Py4J works with
          an independent Python process so full numpy support)
        Does the library handle memory management (e.g. garbage
          collection on objects shared between Python and Java
          processes)?
        What is the threading model between Java/Python
        Related to threading, Is there support for indirect
          recursive calls between Java & Python? (Our recent work
          involved adding this feature to Py4J)
        Is remote support possible or on the roadmap?
        Does the library pass data and objects by call-by-ref or
          call-by-value (i.e. copying) or a combination of both.
        Is there evidence of a good testsuite for the
          functionality?
        Performance!!!! (this is a whole huge area in itself, so
          depends what you are optimizing for e.g individual calls vs
          batch, in due course we would like to formally measure &
          publish some figures for the solutions we are working on)
        
        
        Non-technical considerations:
        -------------------------------------
        What is the project licensing?
        Are there any platform support limitations (We need to
          support Windows, Mac, Linux)
        What evidence is there the project is alive and thriving
          (look for releases in last 6 months, activity on forums)
        Look for evidence of adoption (e.g Py4J is used by Apache
          Spark for PySpark, they also make contributions)
        Look for other evidence of sustainability (any funding,
          companies, institutions involved)
        Is the project open to external contributions?
        Is there a roadmap?
        Is there sufficient documentation?
        
        
        There was no one perfect solution, but Py4J fulfilled many
          of the requirements and we are working on filling in the
          gaps. 
        Let me know if any questions, otherwise I will (separately)
          send out an update on our Py4J activities especially after our
          productive week at EclipseCon France last week. 
        
        
        Regards,
        Tracy