Over the past decade, the availability of large samples of galaxies has allowed us to build a coherent picture of how galaxies form and evolve, now reproduced with remarkable realism by modern cosmological simulations. Yet, many of the key physical processes that shape galaxies remain only partially understood—as revealed, for instance, by the surprising discoveries of the James Webb Space Telescope at cosmic dawn. At the same time, the rapid growth in both the quality and complexity of astronomical data, together with advances in artificial intelligence, is opening new avenues to explore the physics of galaxy formation. In this talk, after a brief introduction, I will present some of our group’s recent efforts to extract physical insight from large multi-wavelength surveys and to bridge the gap between cosmological simulations and observations using modern AI techniques.