France: PhD Position, Neural Implicit Representation/Operator Learning

[administrator]

Full Announcement:
https://jobs.inria.fr/public/classic/en/offres/2024-07293

Applications are welcome at the Inria Strasbourg Antenna, located in the city center of Strasbourg, at the University of Strasbourg, for a PhD position (3 years duration), to start in September 2024 or after a 4 to 6 month Master internship. This is a multi-partner collaborative project with the Culham Centre for Fusion Energy, located close to Oxford, Uniten Kingdom; frequent visits to Culham are planned. The project will also be immersed in a stimulating ecosystem of researchers and students as part of PEPR PDE-AI and Numpex (exascale computing). Existing links with Nvidia and Caltech at UKAEA will also be relevant to the PhD project. The goal of the project is to study neural network-based approaches solving multi-scale problems. We will investigate approaches that maintain accuracy and stability over long times on general geometries. To that end, a balanced approach combining traditional methods and innovative techniques related to machine learning (Neural Galerkin, Neural Operators) will be introduced to solve complex problems in the physical sciences. In particular, applications include fluid dynamics and magnetohydrodynamics, with turbulence and convective mixing, with potential applications to engineering and fusion.

The PhD position is open to students curently enrolled in an MSc in Applied mathematics or Scientific computing or connecting disciplines but with a strong mathematical background. The candidate is expected to have a solid knowledge in numerical methods for the solution of PDEs, basic programming skills and a minimal knowledge of Machine Learning.

Application must be sent online via the official Inria webpage above. The application package needs to contain a CV, a motivation letter, copies of relevant course transcripts, and the contacts of at least two academic references.