Postdoc / research assistant in scientific machine learning

We build physics-informed neural networks and neural operators (DeepONet/FNO-type) as fast surrogates for electrochemical simulations: lithium-ion batteries and solid-oxide electrolysers, where the full PDE models are too slow for state estimation, lifetime prediction and real-time control. The models are trained on simulation and laboratory data and deployed on GPU and edge hardware. Our latest result on mechanistically guided residual learning was published in Nature Communications (January 2026, with MIT).

You have an MSc or PhD in computer science/AI or AI for engineering, are fluent in PyTorch, use Claude Code professionally in your daily work, and have – or are ready to acquire quickly – a working understanding of the physics-based models we accelerate.

We offer an NVIDIA B300 cluster and RTX-4090 workstations, NVIDIA edge and NXP NPU platforms, battery test labs, and collaboration with Stanford, MIT, Imperial College, Chalmers and DTU.

CV and a short motivation to ret@et.aau.dk. Open until filled; decision aimed for mid-December.