Gen.jl

Why we included this project

Gen.jl is a probabilistic programming system embedded in Julia, built for problems where the real challenge is representing uncertainty, not just learning a pattern. Instead of wiring together off-the-shelf estimators, you describe a generative model and then pick the inference algorithm that fits your data and constraints. The programmable inference is what makes it stand out: when the default recipes are too slow or too coarse, you can supply your own proposals or variational families, and you can also swap in a different MCMC kernel without reworking the math. That kind of flexibility is valuable in scientific work, where models rarely match textbook examples, and it makes Gen.jl useful for researchers in Bayesian statistics and robotics, as well as computational science more broadly. If your team already works in Julia and wants modeling and inference to be editable pieces rather than a black box, this is worth trying on your own data.