Within the next couple of years, we will witness the arrival of two new space observatories which will revolutionise the field of galaxy formation. The James Webb Space Telescope (JWST) which launch is planed for December 2021, will unveil, for the first time, the optical rest-frame morphologies of galaxies in the very early universe (z>3). The Euclid survey (launch February 2023), will observe ~35% of the visible sky at a spatial resolution comparable to the one currently delivered by the Hubble Space Telescope. This project revolves around data from these two upcoming facilities - linked to top-of-its-class cosmological simulations - to advance in our understanding of how galaxies acquire their morphological properties over cosmic time from a novel data-driven perspective. We will use state-of-the art supervised and self-supervised machine learning to (1) unveil the non-parametric Star Formation Histories of millions of galaxies, (2) establish new constraints on the role of SN feedback in regulating star formation in high redshift disks (3) provide the first morphological description of galaxies at z>3 and (3) deliver an unprecedentedly detailed census of the amount of accreted stars in galaxies. The project is part of on-going effort to build a new cross-disciplinary Machine Learning Group at the IAC.