The purpose of this task is to harness the photonics components developed in WP3 and WP4 to realize fast and efficient photonics processor capable of performing tensor operations (multiply and accumulate, matrix vector multiplication etc.) with characteristics in terms of footprint, data rate and energy usage beyond the state of the art. Neural networks are the algorithms used by AI protocols, and tensor operations constitute the core of artificial neural networks, each layer of which can be modelized as a matrix vector multiplication followed by a nonlinear transfer function. The four sub-tasks will demonstrate four different approaches to realize fully photonic AI architectures.