Leon’s research has been primarily focussed on the ATLAS trigger system with an emphasis on both the current system and future upgrade. He has been part of the new and unconventional trigger level analysis (TLA) team in ATLAS, searching for dark matter candidates in under-constrained low mass search regions. The TLA programme at ATLAS has increased sensitivity to low mass signatures by reducing the event size and consequently allowing an increased trigger event rate. Searching for Dijet final states in the ongoing LHC Run 3 searches the TLA will be the first time a novel machine learning algorithm for fast jet tagging/classification is used in a physics analysis.
The ATLAS trigger system will undergo a comprehensive upgrade in advance of the HL-LHC programme. In the HL-LHC era there will be even stricter latency and bandwidth constraints, and Leon has helped to develop fast machine learning (ML) algorithms and evaluate their performance in the future trigger system. Typically, hadronic physics reconstruction requires aggregating collimated sprays of particles, named jets. An object detection convolutional neural network (CNN) for jet finding in the ATLAS calorimeter was assessed and compared to the current algorithms. The modified computer vision architecture was found to be approximately one order of magnitude faster on heterogeneous architectures.
