Great idea, hope that will lead to more practical usage of deep learning. Any plans of building one for datasets as well? Something like Kaggle's collection, but used in research papers.