With ongoing and future experiments, we are set to enter a more data-driven era in astronomy and astrophysics. To optimally learn the Universe from low to high redshift I advocate besides new observational techniques for the application of modern machine learning. For example in 3D, tomography of line intensity such as the 21cm line of hydrogen can both teach about properties of sources and gaseous media between. In this talk I first showcase the use of networks that are tailored for tomography to directly infer dark matter and astrophysical properties. I compare network models and highlight how a comparably simple 3D architecture (3D-21cmPIE-Net) that mirrors the data structure performs best. I present well-interpretable gradient-based saliency and discuss robustness against foregrounds and systematics via transfer learning. I complement these findings with a discussion of lower redshift results for the recent SKA Data Challenge, where hydrogen sources were to be detected and characterised in a large (TB) 3D cube. I will highlight my team'92s lessons-learned; our networks performed especially well when asked to characterise flux and size of sources bright in 21cm. Finally, moving from 3D to 1D, for the classification infrastructure group of the new ESO workhorse 4MOST (4-metre Multi-Object Spectroscopic Telescope), I propose an object classification layer to efficiently group the ~40,000 spectra per night (40 million in total) the instrument will collect.