As an intern, you will work on the Fast Physics project, where the main objective is to research, improve, and extend AI models that can act as fast surrogate models for high-fidelity ship-performance simulations. The internship is primarily focused on artificial intelligence and scientific machine learning, with CFD data used as the learning target and validation basis.
Rather than running time-consuming physics-based simulations for every design iteration, we develop AI architectures that learn from vessel geometries, operating conditions, and simulation outputs to estimate quantities such as resistance, pressure distributions, and flow fields. A central topic is the exploration of recent Physics Transformer models and neural-operator architectures, and how these can be adapted to complex maritime geometries.
You will contribute to improving our current Physics Transformer model and architecture by benchmarking recent research, designing model improvements, running training experiments, and validating performance across different hull forms, operating conditions, and simulation fidelities. The outcome should be a stronger AI model prototype and a clear research contribution on how transformer-based physics models can support fast simulation and early-stage design exploration. The assignment can be a thesis/graduate internship and could start from September onwards.