With the increase of large spectroscopic surveys like Gaia, APOGEE, and the California-Kepler Survey, a new class of data-driven spectrum models are possible. Unlike their predecessors, which rely on physical laws and radiative transfer calculations, data-driven models are trained on large sets of real spectra to generate models that closely resemble the data. I've trained data-driven models on datasets from Gaia-DR3 and the California-Kepler Survey. The accuracy of these models enables the discovery of spectroscopic anomalies like active stars and binaries.
View the publications for this work here and here, and a video of me presenting some of our findings here. I also published an open-source package that uses our models to detect anomalous stars in Gaia.