RPyC seems to transfer int, long, bool, str, float, unicode, bytes, slice, complex, tuple (of simple types), forzenset (of simple types) as well as the following singletons: None, NotImplemented, and Ellipsis
(from the RPyC manual), so as you can send unicode, you can send code as a string, and eval it server side.
Note that if you do so, I'll be able to send things like os.unlink(sys.argv[0])
on your server, or this kind of stuff you may not want. In other words allowing the server to run arbitrary code is a huge security breach.
So if you're in a server to server case, it's a go but I don't think you really need it, if you completely trust both sides, but if it's a client-server application, it's a no go.
If you decide to go this way, eval may not be the right choice, opening your server far too much. You have explore some ways to restrict possibilities, according to what you want to allow your bunch of logic
to do, search about "python eval security", you'll find some interesting threads.