Large galaxy surveys aim to unravel the nature of dark energy and dark matter with precise and accurate measurements of cosmic structures. This goal can be achieved by joint analyses of multiple cosmic structure tracers, including galaxies, shears, and galaxy clusters. However, practical applications are hindered by two significant challenges: a lack of realistic mocks that consistently model all three different tracers and enormous computational demands on likelihood inferences. In this talk, I will show two recent efforts at solving these two obstacles. I will show the Cardinal simulation, featuring unparalleled accuracy in color-dependent clustering and cluster abundance. I will then describe the new machine learning-based likelihood inference tool that speeds up the survey analyses by a factor of 50. Finally, I will show some prospects of the combined analyses on constraining the impact of baryonic feedback on cosmic structures.