PhD position - Parameter inference for PDE models using PINNs

PhD outline The PhD will develop a Fokker-Planck PDE-based framework for parameter inference in SDE models using destructive and sparse biological measurements. The work will consist of: (i) formulating inference problems under partial or noisy observations; (ii) developing PINN-based metamodels to solve Fokker-Planck PDEs efficiently in moderate-to-high dimensions; (iii) integrating these metamodels into parameter estimation, either directly or through hybrid strategies (e.g., PINN– MCMC); (iv) applying and validating the methodology on real datasets from plant–pathogen interactions, poultry host–microbiota– pathogen dynamics, and biofilm microbiota–pathogen systems. The doctoral candidate will have access to a dedicated laptop, local cluster computing resources, and GPU facilities at the Lab-IA for large-scale computations, as well as real datasets already available from the SMILE (ISA), MiMoSa (ISP), and B3D (Micalis) teams. Candidate Skills The PhD candidate should possess strong mathematical modelling skills (SDEs, PDEs), very good programming experience in Python, and ideally some familiarity with machine learning. A strong interest in interdisciplinary biological applications and collaborative research is essential.