Postdoctoral Researcher (24 months) – Hybrid Mechanistic & AI Modeling of Microbiomes

You will build the computational backbone of a multi-scale modeling framework that links how individual bacterial strains metabolize plant matrices up to how whole microbial communities behave, so we can predict and steer which combinations of strains keep a targeted strain alive and active through fermentation. Concretely, you will: Construct and curate genome-scale metabolic models (GEMs) for food-fermentation strains, using deep-learning-based reconstruction (e.g., DNNGIOR) and curated metabolic modules; Run static and dynamic flux balance analyses across thousands of simulated microbial consortia to prioritize strain combinations that support a targeted strain; Embed the resulting metabolic descriptors into a hybrid deep-learning model (compositional Neural ODEs) constrained by consumer–resource ecological equations; Combine scarce high-resolution experimental data with large public datasets and drive training with synthetic data from model forward simulations; Parameterize dynamic models with Physics-Informed Neural Networks (PINNs) to enforce biophysical constraints; Integrate an open multi-omics dataset with a validated predictive model to feed the project's digital-twin fermentation platform. You will work at the interface of simulation and experiment, interacting closely with the wet-lab and digital-twin teams, with access to the MIGALE and Jean Zay computing clusters.