A new machine learning workflow helps decode low-loss EELS data, turning noisy nanoscale spectra into spatial maps of the optical resonances that shape next-generation hybrid nanophotonic materials.