I'd recommend doing a startup and/or side project involving interesting functions of massive datasets.
Technology-wise, at least consider...
- learning SML or OCaml or Erlang
- learning Python
- going deeper into Python. It's a fantastic way to express oneself. It beautifully offers (generators, coroutines, multithreading, subprocesses, and networking) for piping and distributing your computation, a flexible blend of functional and imperative, and its syntax is the most conducive to elegance, simplicity, and rereadability of anything I've tried.
...then building the core algorithms in C/C++ with SWIG wrappers, the tip of the iceberg with Django, and everything else in pure Python.
A must-read: "Generator Tricks For Systems Programmers"