The current concordance (LCDM) cosmological model provides a robust framework for understanding large-scale structure, yet it faces persistent challenges on scales smaller than 1 Mpc. Discrepancies such as the core/cusp and missing satellite problems, along with enigmatic galactic scaling relations like the Radial Acceleration Relation (RAR), suggest that our understanding of dark matter remains incomplete. This talk explores the nature of dark matter across cosmic scales, from individual galaxies to massive clusters.
The first part of this work focuses on observational tests of the LCDM paradigm. By analyzing the constancy of dark matter halo surface density and the RAR within galaxy groups and clusters, we compare empirical data against multi-wavelength mock catalogs.
The second part of the presentation transitions to the next frontier of digital sky surveys: the application of deep learning to uncover hidden gravitational lenses. We detail a machine-learning-driven search for galaxy-galaxy strong lenses in the infrared sky, which has been tailored to identify rare red-red lenses at high redshifts. These systems serve as powerful natural telescopes, offering a unique pathway to probe the fainter quiescent galaxies at the higher redshifts. By bridging small-scale observational constraints with deep learning, this research provides new insights into the fundamental properties of the invisible universe.