The Simons Collaboration, Learning the Universe, is working toward one lofty goal: how can we measure the initial conditions of our universe and the physical laws that govern it, given what we're able to observe? How do we get to this using cutting-edge results and progress in: simulations on the biggest and smallest scales, galaxy formation and evolution models, statistical inference and forward modeling, and synthetic observations?

I'm helping us get there by showing this is possible on a small scale, directly on a real observed galaxy sample. CAMELS-SAM is at the heart of the ambitious "LtU Connections" project to build a first-of-its-kind pipeline for simulation based inference of cosmology from the SDSS Main Galaxy Sample. I'm connecting the knowledge and methods of the training set generation group, the implicit likelihood inference group, and the synthetic observations group to do this.

See the figure below for a schematic of how I'm doing this, and check back in often for updates! As of mid-October 2026, I'm working with several folks in the synthetic observations group who are experts of dust modeling and using the Synthesizer code to incorporate the Sommovigo & Bartlett dust model (read more here) to CAMELS-SAM galaxies. This means that the dust modeling will respond to the galaxy formation, too, as has been seen in high resolution simulation projects the team has done -- meaning that our inference can learn from this information too! I'm also leveraging the work I did with a great student to rescale the cosmology of the CAMELS-SAM merger trees directly to add thousands of more galaxy catalogs sampling more cosmologies to our data. Exciting times!