DTU Compute invites applications for a Postdoc in Bayesian Deep Learning within the Department of Mathematics and Computer Science. The position focuses on understanding why current Bayesian deep learning methods fail at scale and developing robust, scalable Bayesian approximations for deep neural networks. The Postdoc will work in a collaborative research group led by Prof. Søren Hauberg, emphasizing theoretical insight, numerical methods, and practical machine learning. Research topics include approximate Bayesian inference, differential geometry, large-scale computation, and modern ML frameworks such as JAX. Results are expected to be published at leading international machine learning venues.