I wish to use this computing resource for my PhD in general for which I am working on two specific projects now. Both projects include the construction of so called "String vacua" using machine learning tools. More precisely, I will be using Reinforcement Learning methods (DQNs) and Generative Models (Discrete Diffusion and/or Bayesian Flow Models) in order to sample string vacua. The reason for my interest in those configurations is the following: In order to be consistent, String Theory requires the existence of ten space-time dimension while our universe clearly consists of only four (three spacial dimensions and one time direction). In order to deal with this discrepancy, one aims to compactify String Theory on some six-dimensional space that is wrapped up such that we cannot detect it at low energies. While this topic requires a strong understanding of the relevant mathematical and physical concepts, there remains another problem that cannot be tackled by understanding alone: the space of possible solutions obtained by such compactifications (string vacua) is of such high order that checking solutions by hand becomes intractable, if not impossible. Hence one might want to use tools such as Reinforcement Learning to scan through the space of possible constructions. Choosing a certain six-dimensional space together with other ingredients has direct implications on the resulting physics at low energies. Hence, whether a model is realistic or not can be used as a reward in the context of Reinforcement Learning, teaching the agent to find good solutions. Another way realistic solutions can be constructed, consists of using Generative Models, mostly known for their application in image generation. Such model could be trained on several known solutions to then be used to sample new unknown solutions. This application of Generative Models in String Compactification seems to be a rather uncharted area and is hence an ideal candidate for new projects.