This proposal is made in connection to a Formas-funded project with the same title (grant no 2019-01295), granted for the period 2020-2022. In several earlier and ongoing projects, we have developed advanced methods for quantitative microscopy and measurements of mass transport, image segmentation techniques for characterization of material microstructures, and models to generate virtual materials to explore properties and establish microstructure-mass transport property relationships. Learning how to predict properties faster and being able to tailor-make a material for a specific purpose easier will lead to reduced ecological footprint an financial savings for these industries. The aim of this project is to develop new machine learning-based methods for estimation of local diffusion properties from microscopy data, volumetric image segmentation of focused ion beam scanning electron microscopy data, prediction of effective diffusivity and fluid permeability from microstructural geometry, analysis of small angle X-ray scattering data, and prediction of thickness and crystalline phase from electron diffraction data. These projects all have in common that existing methods can be vastly improved upon or accelerated by utilizing the advances of recent years in the machine learning field, and the incorporation of these new techniques will lead to valuable tools and new insights on materials design and properties.