Reservoir computing (RC) is a promising paradigm for physical artificial neural networks. Because RC uses a fixed network, it lends itself well to realization in a variety of physical systems. The simple and quick training method involved also reduces energy and time costs associated with neural networks. Reservoir computing in optical systems is emerging as a primary modality due to the large dimensionality and parallelism, speed of operation and maturity of photonic technologies. Harnessing basic properties of quantum optics for RC grants advantages from efficient processing of quantum information to improvements in computational power which can scale without the increasing the physical size or complexity of the system. Here we will discuss a platform for RC using linear optical networks and photon-number resolving detectors and we will show how a versatile device for quantum machine learning can be built from off-the-shelf components.