Scaling AI for Science

AI for Science only matters at the scale where discoveries happen. The lab takes machine learning and computer vision from prototypes to production on real physics data: JUNO solar- and reactor-neutrino pipelines, detector-response modelling, and reconstruction that runs within the experiment’s computing budget.

The standard is scientific: a system is intelligent when its claims survive the scientific method — define observables, run controlled experiments, build effective theory, and validate by prediction. The lab applies this standard both to physics analyses and to the AI systems that help produce them.