The MASH project is a three year European initiative which started early in 2010. Its main objective is to develop new tools and methods to help the collaborative design of very large families of features extractors for machine learning. This is investigated through the development of an open web platform which allows to submit implementations of feature extractors, browse extractors already contributed to the system, launch experiments, and analyze previous experimental results to focus the design on the identified weakness of the system. Performance of the designed learning architectures are evaluated on image recognition, object detection, and goal planning. The former targeting both a video-game-like 3d simulator and a real robot controlled remotely. We want mainly to demonstrate the features of the platform to the NIPS attendees, and show both the development tools, and the applications we have already implemented.