Abstract: This seminar will discuss NuGraph2, a Graph Neural Network (GNN) for reconstruction of liquid argon time projection chamber (LArTPC) data. The network leverages a multi-head attention message passing mechanism to classify detector hits according to the particle type that produced them. Performance results will be presented based on publicly available samples from MicroBooNE, including both physics performance metrics and computational metrics for training and inference on CPU and GPU. Also discussed will be the open source NuML toolkit, a set of packages providing generalisable and extensible algorithms designed to handle many of the boilerplate technical aspects of training a neural network in particle physics.