Following the transition to multi-core, we will observe another fundamental transition towards communication-centric architectures and systems, in which communication costs will dominate in terms of energy and time. This transformation will percolate up through the software stacks, resulting in a complexity that cannot be exposed to the user. Key for future research is closing this semantic gap between architecture and user. My objectives include increasing the performance, energy efficiency and usability of parallel computing systems, with research interests on hard- and software co-design, programming models, heterogeneity and energy-awareness. While I understand myself as computer architect, my research targets a variety of applications, ranging from machine learning, high-performance analytics and high-performance computing to embedded devices and cyber-physical systems.