Job offer

This research position aims at developing methods for inferring the internal structural parameters of lipid nanoparticles from small-angle X-ray scattering (SAXS) measurements, drawing on parsimony-based and dictionary learning approaches. The intrinsically ill-posed and non-unique nature of SAXS signal inversion, combined with the great diversity of morphologies observed, leads to the development of representations capable of decomposing the scattering profiles into dictionaries of elementary patterns reflecting the different structural organisations of the nanoparticles. This approach will provide a naturally interpretable framework for linking dictionary-based representation to the structural characteristics of nanoparticles.

The core of the work will involve designing dictionary-learning methods tailored to SAXS data, enabling the automatic extraction of atoms representative of the scattering signatures and the reconstruction of experimental profiles using parsimonious combinations, whether linear or non-linear. Particular attention will be paid to the integration of physical and biophysical constraints into the learning process, in order to ensure that the representations obtained remain consistent with the known properties of lipid nanoparticles.

Strategies involving hierarchical, structured or multimodal dictionaries may, in particular, be investigated in order to represent different scales of internal organisation. Another major challenge lies in the limited availability of annotated data. The candidate will therefore explore semi-supervised or self-supervised dictionary learning strategies, combining numerical simulations, experimental data and a priori knowledge derived from the physics of diffusion. These approaches must also be robust to measurement noise, instrumental variations and changes in domain induced by different pharmaceutical formulations, in order to ensure good generalisation ability across data acquired in a variety of contexts.

PhD or equivalent, Minimum qualification required: 8 years of higher education: PhD in signal and image processing / statistical learning / artificial intelligence / inverse problems or related fields.

Core competencies:

• Solid theoretical and practical knowledge of signal and image processing, statistical learning and optimisation methods

• Experience in developing and solving inverse problems, ideally in the context of scientific or experimental data

• Proficiency in scientific programming, particularly in Python, as well as in data analysis and processing tools

• Ability to develop, implement and evaluate innovative methods in a research context

• Ability to work independently on a scientific project, from problem formulation through to the validation of the approaches developed

• Ability to work in an interdisciplinary environment, interacting with experts from applied fields

• Skills in scientific communication, writing papers and presenting results Specific Requirements

Remote work 2 days a week is possible

Languages ENGLISH