Task leads: Nikos Konstantinidis (UCL), David Miller (Chicago)

The Level-0 Global Trigger (L0Global) is a new subsystem, which will execute offline-like reconstruction algorithms on full-granularity calorimeter data in real time at 40 MHz, with latency of a few microseconds and data throughput of 50 Tbps. Novel Machine Learning based reconstruction and feature extraction algorithms need to be developed to extend the physics potential of the experiment. We will develop a common framework for optimizing such algorithms in four main areas of the Global Trigger: electrons/photons, taus, jets, and multi-object, full-event reconstruction; as well as for the preprocessing of the calorimeter data inputs. The ultimate goal is to develop and optimize these algorithms all the way to full firmware implementation and integration into the Global Trigger system to augment or replace the baseline algorithms during Run 4 of the LHC.