JEM: An AI-based engine workflow to predict simulation’s execution time on HPC cluster

Within the automotive industry, numerous simulations are essential for accurately modeling a vehicle’s behavior and its components. However, these simulations entail an unpredictable duration for execution. Providing estimates for the required time enables users to organize their workload better and optimize resource utilization. Furthermore, the characteristics of submitted simulations can evolve, influenced by factors such as the component type, final product, car model, and other variables. In this dynamic context, staying abreast of these changes becomes imperative.This demo paper introduces the JEM tool, designed to accomplish two key tasks: (1) estimating the runtime of simulations and (2) allowing data scientists to monitor changes in data distribution and triggering model retraining when necessary. This tool results from a collaborative machine learning research project involving Iveco Group, Doit Systems, and Politecnico di Torino. Integrated into the Iveco group’s HPC system, JEM facilitates resource allocation.The recorded demo of the tool provides a demonstration of how users can interact effectively with the JEM tool.