The goal of this project is to research and evaluate architectures and runtimes for scalable and energy efficient next-generation multicores and accelerators. High Performance Computing and AI are limited by bottlenecks in resource sharing, communication and energy. In this project we will evaluate advanced computing architectures, as well as programming models, compilation techniques and runtimes to manage and schedule tasks to heterogeneous hardware with the overall goals of high performance and reducing energy consumption in the domains of HPC and AI.