EFfective Field TheORy SurrogaTe: A Cosmological Emulator (github.com)

Effort.jl is an open-source surrogate emulator for the Effective Field Theory (EFT) of Large Scale Structure (LSS), released by Marco Bonici and collaborators. The code is written in Julia and is explicitly designed to be fast and differentiable, providing rapid evaluations of EFT-based model predictions used in “full-shape” galaxy clustering analyses. The repository is free to use with a citation to the release paper and has already been applied in several recent studies, including DESI DR1 analyses and work on HOD‑informed priors and nuisance-parameter reparameterizations. The significance for the AI/ML and cosmology communities lies in the combination of speed and differentiability: fast emulation reduces the computational cost of likelihood evaluations, while automatic differentiation enables gradient-based inference (e.g., HMC or variational methods), end-to-end optimization, and sensitivity analyses. Practically, Effort.jl has been used to build HOD-informed priors, reduce sensitivity to nuisance priors via non-linear reparameterizations, and select galaxy samples with fewer Fingers-of-God—demonstrating improved robustness and tighter constraints in cosmological parameter estimation. For researchers doing large-scale inference or developing ML-driven pipelines for cosmology, Effort.jl offers a plug‑in, production-ready emulator that accelerates EFT-based modeling and enables modern, gradient-aware inference workflows.