Computational models of biosystems are getting increasingly  adopted by industry to reduce costs of product design and to increase productivity and efficiency. New technoglogies developed in pharmaceutical, biotech, and biomedical companies require more and more usage of in-silico modeling and  "Digital Twins".  A whole variety of so-called mechanistic models having a high potential to simulate complex processes (DEM,CFD, ..),  has gained a lot of popularity over the last 30 years. These modeling techniques can offer great insight, but despite the computational means, a fluent adoption by industries is often impaired by their long computation times, the former requiring most often "accurate enough",  but fast predictions. Over the past 10 years, there has been a growing interest and usage of hybrid models and surrogate models. These  models need to be "trained" to reproduce the result of mechanistic models, yielding new models with a lower execution time and an acceptable accuracy. I will aim at developing techniques to construct hybrid and surrogates for mechanistic models, putting the emphasis on  "particulate systems" (granules, powders, multi-cellular systems,..), which are of daily importance in pharma and biotech. I intend to hire a PhD student to start with this project.