In this talk I will present how I use cosmological simulations and high-performance computing ressources to predict and understand gravitational wave events. We model compact object binary populations based on the combination of binary population synthesis models and cosmological simulations from the FIRE project. The Feedback In Realistic Environments project develops cosmological simulations of galaxy formation that directly resolve the interstellar medium of individual galaxies while capturing their cosmological environment. We specifically use the simulations from Milky-Way like galaxies to model compact binaries. We are effectively building mock catalogues of white dwarf binaries, containing close to a billion sources, which will be used to analyse and predict gravitational wave (LISA) and electromagnetic surveys (Gaia, LSST). I will present our future project with the large volume FIREBOX simulation recently completed on Mare Nostrum. Post-procesing this simulation will yield compact object merger rates directly comparable with the next Virgo/LIGO runs. I will emphasize the specific challenges related to the volume of the outputs and the generation of mock catalogues.