Stage III cosmological surveys have imaged tens of millions of galaxies; Stage IV surveys now beginning operations will yield billions. AI-accelerated workflows are making robust inference of redshifts and physical properties feasible at these scales. This project will apply their pop-cosmos framework to infer astrophysical properties for 41 million galaxies from the landmark Kilo Degree Survey (KiDS). The pop-cosmos framework combines a neural emulator for stellar population synthesis, a diffusion model prior calibrated on deep COSMOS data, and a GPU-optimized Bayesian inference pipeline. This unprecedented analysis will enable precise and robust KiDS cosmology and establish the methodology for billion-galaxy catalogues from the Vera Rubin Observatory's Legacy Survey of Space and Time (LSST).
Hiranya Peiris, University of Cambridge, United Kingdom