Katja Seeliger obtained an undergraduate in computer science at Bauhaus University in Weimar. A research project at Macoto Kikuchis lab at Osaka University introduced her to the idea of irreducible emergent computations in physics and biology. This sparked a deep interest in the computational aspects of nature, and within neural systems in particular.

She proceeded with a master's in computational neuroscience at the Bernstein Center for Computational Neuroscience Berlin – specialising on the similarities between the then emerging modern neural networks and biological visual cognition in the lab of Shinji Nishimoto. She obtained a PhD in this field at the artificial intelligence department of Donders Institute at Radboud University Nijmegen, under the supervision of Marcel van Gerven.

During her PhD, she uncovered how the hierarchy of neural network layers is followed in time with a large-scale MEG recording. She pioneered the use of generative adversarial networks for reconstructing natural images from brain activity. She recorded the largest sensory fMRI dataset in a single individual to-date, and used it to spearhead training neural networks end-to-end on this data, creating an interpretable digital copy of this individuals’ visual information processing across large parts of his visual system.

In her postdoctoral research, she is using similar end-to-end-trained models to study the higher visual system, uncovering new insights into properties the higher visual system responds to. She is also working on pushing the field of neuroAI by recordings of new large-scale neuroimaging datasets.