Reservoir computing is a paradigm from the area of machine learning, which is shown to have good performance on temporal dynamic problems. In this project I am targeting the breakthrough of building prototypes of such photonic reservoir information processing platforms, with a special emphasis on ultra-low power consumption. These platforms will be integrated on a chip, and will perform machine-learning computations in industrially-relevant applications (medical image recognition, telecommunication). They will be able to do this at very high speed and ultra-low power consumption, which surpasses what is possible with other (typically electronic) information processing approaches.