The ACROSS platform aims at providing access to the software stack that enables application workflows mixing large numerical simulations, machine learning (including deep learning) training/inference operations and big-data analytics to be efficiently executed on HPC resources, as well as taking advantage from diverse hardware accelerators including neuromorphic-like architectures.

As such, the ACROSS software stack integrates software components that allow the user to easily define the application workflows, as well as matching with the most suitable execution environment. The ACROSS software stack is enriched with smart mechanisms to acquire HPC/cloud resources on-demand and effectively manage them, easily enabling co-location of jobs on heterogeneous compute resources.

On top of this stack, a simple rather flexible command line interface (CLI), along with a rich set of rest API to create cloud deployable templates, defines the ACROSS API, which provides user access to the compute resources.