Hi, i'm Timon.
i'm a computer science master's student at TUM, working on machine learning — currently on world models and interpretability.
About
i'm doing my M.Sc. in Computer Science at TUM, specialising in Machine Learning and Analytics, after a B.Sc. at the same place with a minor in mathematics. My bachelor's thesis was on subspace diffusion for channel estimation — applied generative modelling.
Some of my favorite courses include: introduction to deep learning, natural language processing, and a master's seminar on advanced topics in vision–language models, protein prediction for computer scientists and efficient algorithms and data structures.
Projects
A JEPA world model that fuses four synchronised camera views without reconstruction or contrastive terms. Multi-view fusion makes physical state markedly more legible in the latent space — and, honestly reported, buys nothing measurable downstream.
PyTorch · JEPA · VGGT · OGBench · MuJoCo