The project involves research on new, high performance multi-chiplet chip architectures aiming to support more efficiently data-intensive workloads. The volumes of data needed by modern processors is rapidly growing limiting their performance and energy efficiency. That is, firstly, because new emerging applications become increasingly more data-intensive and, secondly, because the number of cores and specialized accelerators integrated on a processor chip keeps increasing, requiring rapidly growing data rates to process. The current processor-centric computing systems are not able to handle such large data rates efficiently. Their performance is limited because their cores remain idle waiting for data. Moreover, large data volumes trigger unnecessary data movement which wastes energy. The proposition of this research is to rethink the way computing systems handle data and make it data-centric rather than processor centric. Considering a multi-chiplet processor system, the plan is to provide mechanisms that involve all system aspects, namely, processing, memory and interconnects to tackle the following two challenges: (i) minimize the data movement in order to maximize processing near data, and (ii) optimize data transfers when they are necessary.