As next generation cosmological surveys reach unprecedented depths over large cosmic volumes, hydrodynamical uncertainties in observed statistics are becoming an increasing concern. Current techniques to marginalize out these uncertainties require running suites of large volume simulations, requiring significant computational resources. In this talk, I will discuss an alternative relying on constructing a variational auto-encoder for a hydrodynamical field conditioned on the corresponding dark matter field. Once trained on one hydrodynamical simulation, we are able to probabilistically map separate dark matter only simulations to full hydrodynamical outputs. By sampling over the latent space of the VAE, we can generate posterior samples and study the variance of the mapping. We find that our reconstructed field provides an accurate representation of the target hydrodynamical field as well as a reasonable variance estimate. This approach has promise for the rapid generation of mocks as well as for implementation in a full Bayesian inverse model of observed data.