Speeding up scikit-learn workflow using a high-performance Go proxy (ioloop.io)

interesting. normally I just save my model in sklearn, but I like how this does the caching transparently before even building the model.