Offer Description Applications are invited for an 18-month, full-time postdoctoral research position in the mathematical and statistical foundations of machine
learning, working with Eddie Aamari. The position is supported by the PERSNET project within the French PEPR Artificial Intelligence programme, funded by France 2030. The successful candidate will be based in Paris at the Département de mathématiques et applications (DMA), École normale supérieure – PSL, within the Probability and Statistics team and the ENS-PSL Center for Data Science . The research programme will be developed jointly with the successful candidate, taking into account their expertise and interests within the scientific scope of PERSNET. Possible research directions include: Topological and geometric analysis of neural networks. Transformers, attention dynamics, tokenization and learned representations. Generative modelling on manifolds, stratified spaces and singular data. Diffusion and score-based models, including their connections with curvature, reach and tangent/normal geometry. Drifting models, kernelized transport, gradient flows associated with maximum mean discrepancy and reproducing kernel Hilbert spaces, and nonlocal partial differential equations. RKHS geometry, kernel mean embeddings and geometric statistical testing. The main duties are to conduct original research, prepare scientific publications, present results at seminars and conferences, and participate in the scientific activities of the host team and the PERSNET project. Where to apply E-mail: eddie.aamari@ens.fr Requirements Research Field: Mathematics » Statistics Education Level: PhD or equivalent Skills/Qualifications Applicants should demonstrate: A strong mathematical background in one or more relevant areas, including probability, statistics, geometric inference, optimal transport, partial differential equations, stochastic processes, topological data analysis, kernel methods or mathematical machine learning. The ability to develop rigorous mathematical arguments and pursue original research with increasing independence. The ability to collaborate and communicate research effectively in English. Previous experience with generative models, transformers or numerical experiments is welcome but is not required. Expertise in all the research directions listed above is not expected.
