Speaker
Description
In this talk, I’ll introduce Kosmulator — a new, fast, and transparent MCMC framework for cosmological inference currently in development. Kosmulator allows researchers to swap cosmological models on the fly for fitting theoretical predictions to observational data. Modified-gravity models are implemented as simple Python callables, meaning transitions from ΛCDM to linear or fully non-linear MG variants require only small mathematical changes, with no re-wiring of likelihoods. Under the hood, Kosmulator integrates CLASS, CLIC, and AlterBBN for theory predictions and supports Zeus/EMCEE MCMC engine samplers, vectorised model evaluations, and high-performance MPI parallelisation. It also includes practical tools for modern analysis workflows: dynamic parameter safeguards, automated convergence and autocorrelation checks, and real-time visualisation (trace plots, corner plots, and best-fit overlays).
