The Java machine learning toolkit Weka includes a “knowledge flow” editor which allows users to very quickly assemble data processing “circuits”:

Similarly, the machine learning toolkit “Orange” includes an interface they call “Canvas” for graphically wiring together machine learning components:

The aim of this project would be to implement a similar “graphical programming” interface for either scikit-learn (a general-purpose machine learning library for Python) or PyLearn2 (focussed on efficient GPU implementations of deep neural networks). For example, it might be nice to be able to graphically specify a deep neural network that is then run on the GPU. Perhaps your interface could integrate with the iPython Notebook.