This proposal is made in connection to the three projects below and two research directions. The majority of the research (in terms of compute) concerns diffraction-based methods for characterizing thin, 2D materials, specificaly position averaged convergent beam electron diffraction (PACBED). This is a relatively new technique for characterizing materials and in great need of method development concerning which methods to use and how to optimize experimental parameters for the best parameter prediction, etc. We have almost finished a first project on predicting deformation parameters such as strain, rotation, and sample tilt, to be submitted soon. We are also working on prediction of phase (crystal structure) and thickness (number of atomic layers), as well as generalizing the almost finished project to more complex mechanical deformations (e.g. deformation in 3D). The projects concern both regression and classification utilizing convolutional neural networks. We are investigating predictions for many different experimental/simulation parameters to optimize the method, resulting in heavy computations both from a data generation (performed on other resources) and an ML training standpoint. This part also requires substantial storage. The smaller part of the research is machine learning-based models for analysis of small angle X-ray scattering (SAXS) experiments. SAXS is a technique complementary to imaging for characterization of materials and molecules at small spatial scales. The field is dominated by fairly simple analytical models, but we are working on complex spatial models for e.g. nanoparticles with non-standard geometries. A key insight, which is new in the field, is that the SAXS signal can me computed using an ML-based surrogate which can then be combined in a superpositions which very acurately can model distributions of properties/parameters, such as size distributions, shape distributions, electron densities, etc. This can lead to considerable speed-ups for modelling SAXS signals for entire complex systems. We intend to finish the investigation for this system as well as developing similar approaches for accelerating data analysis for other systems. This part does not require any additional, non-default storage. The project as a whole involves both experimental specialists from the Eva Olsson group at Physics at Chalmers, and AI expertise from the Mathematical Sciences and Electrical Engineering departments.