22 Postdoctoral Fellowship at Inria, France

Atlantis is  a joint project-team  between Inria and  the Jean-Alexandre Dieudonné Mathematics Laboratory at  Université Côte d’Azur. The team  gathers applied mathematicians and  computational scientists who are collaboratively undertaking  research activities aiming at the design, analysis, development and  application of advanced numerical methods for solving systems of  partial differential equations (PDEs) modelling nanoscale light-matter interaction problems. In this context, the team is  developing  the   DIOGENeS  [https://diogenes.inria.fr/]  software suite,  which  implements  several Discontinuous  Galerkin  (DG)  type methods tailored to the systems  of time- and frequency-domain Maxwell equations  possibly coupled  to  differential  equations modeling  the behaviour of propagation  media at optical frequencies. DIOGENeS also includes a component dedicated to the optimization of geometrical characteristics of nanostructures driven by some performance objective in the contex of inverse design strategies of nanophotonic setups.   DIOGENeS is a unique  numerical   framework  leveraging   the  capabilities   of  DG techniques  for  the simulation  of  multiscale  problems relevant  to nanophotonics and nanoplasmonics.

One important line of research of the team during the last years has been dedicated to improve the capabilities of these numerical tools to produce novel inverse design methodologies for optical metasurfcaes. In the last decade metasurfaces, i.e. 2D arrays of optical nanoantennas with subwavelength size and separation [1] have revolutionized the field of linear optics with the promise to replace bulky and difficult-to-align optical components with ultrathin and flat devices like metagratings, metalenses and metaholograms, which can also implement new functionalities in terms of aberrations correction and arbitrary wavefront shaping. In the recent years, by combining a high-fidelity DG-based fullwave solver in the time-domain [2] with a statistical learning-based global optimization method [3], we have introduced innovative inverse design methodologies for mono-objective optimization of metadeflectors [4], multi-objective optimization of RGB metalenses [5] and robust optimization of metadeflectors [6].

Scientific Machine Learning (SciML) is a relatively new research field bridging machine learning (ML) and scientific computing. Its aim is the development of new methods to solve several kinds of problems, which can be forward solution of PDEs, identification of parameters, or inverse problems.  The methods that are investigated in this context must be robust, scalable, reliable and interpretable. Two main families of methods can be distinguished. On one hand, methods that approximate the solution function, i.e., the mapping from instances of the function variables to the function values, such as with Physics-Informed Neural Networks (PINNS) and their numerous variants. On the other hand, methods that approximate the solution operator, which are generally classified as Neural Operators (NOs). Each of these two families has advantages and drawbacks when one is willing to consider complex PDE models of realistic physical problems. NOs require data, and when that is limited or not available, they are unable to learn the solution operator faithfully. PINNs do not require data but are prone to failure, especially on multi-scale dynamic systems due to optimization challenges. In this postdpctoral  project, we will focus on NOs in the context of time-harmonic electromagnetics wave propagation in heterogenous domains involving irregularly-shaped geometrical features. The overarching goal will be to design NOs that can efficiently deal with the system of time-harmonic Maxwell equations for the complex-valued electric and magnetic fields with different types of boundary conditions and source terms in two- and three-dimensional settings, and data from unstructured mesh-based FEM (Finite Element Method) simulators.  In addiiton, these NOs shall ultimately be capable of generalization over different geometrical characteristics of scattering structures to serve as fast surrogates in inverse design strategies for finding optimal scatterer shapes driven by a performance objective.

In recent years, the number of disease-modifying treatments for Multiple Sclerosis (MS) has augmented significantly (McGinley, Goldschmidt, and Rae-Grant 2021). In particular, highly effective second-line immunosuppressive treatments have become available and the number of first-line treatments has increased. However, these treatments are not without potential adverse effects. It is therefore crucial to prescribe the right treatment to the right patient, and to monitor its effectiveness and safety closely.

Currently, Magnetic Resonance Imaging (MRI) plays a central role in this context. In particular, MRI allows:

– the identification of MS lesions in particular regions of the central nervous system during the first years of the disease;
– the identification of new hyperintense MS lesions between two longitudinal MRI scans i.e. at two different time points.
The two above elements are central, each with their own contribution, to select a patient’s initial treatment as well as to modify the treatment over time.

The Empenn team is one of the leaders of the Primus project. Primus (standing for “Projection in Multiple Sclerosis” (PI: Prof Gilles Edan, Rennes University Hospital)) was granted by the French Ministry of Health in 2022. This project gathers together researchers, faculty members, clinicians and private companies, with the goal of developing a clinical decision support system for Multiple Sclerosis diagnosis and follow-up. One of our contributions is dedicated to the development of methods that allow for detection and segmentation of Multiple Sclerosis lesions from spinal cord MRI images acquired with current clinical protocols. It must be emphasized that MS lesion segmentation in spinal cord is a complex task due to some major challenges such as the size of the anatomical structures of interest (the spinal cord ~ 1cm diameter) and the occurrence of significant artifacts due to motion and respiration. Over the past years, we led several works in this area.

General purpose fault tolerant strategies lead to excessive execution of recovery tasks (re-execution of tasks on failed machines). Therefore, we will investigate how to adapt fault-tolerance techniques to P2P systems by making job scheduling failure-aware (leveraging our previous experience and work with Hadoop clusters [4, 5]) and by enabling checkpoint/restart so that we can roll back execution from the last checkpoint instead of restarting the execution after a failure [6]. We will present a performance model for checkpoint/restart in P2P systems and introduce a scheduling framework that decides when and where to trigger checkpoints and where to restart, and when and where to execute recovery tasks, taking into account failure distribution, data location, and resource heterogeneity. We will also explore how to use P2P storage services (e.g., hive-Disk platform) to store checkpoints and temporary data (e.g., map outputs in MapReduce).

We will study the interplay and correlation between the different factors that contribute to the performance of geo-distributed workflows (i.e., input data, intermediate data, number of iterations, capacity of site, etc). Accordingly, we will design scheduling policies and associated data movement to improve the overall performance, monetary cost, and resource utilization by considering the input location, the network cost and status, intermediate data, and the capacity of the different sites, when scheduling multiple workflows across massively distributed environments.

The aim of the postdoc is the design and study of novel approaches for the development of stochastic geometric numerical integration. Depending on the background of the recruited person, the project will focus on the creation of new stochastic integrators on manifolds, the study of the algebraic and geometric structures underlying stochastic numerics, or the design of robust methods for solving multiscale stochastic dynamics.

The proposed position will focus on the development of grant-free access techniques for IoT applications.

One common communication scenario in IoT applications is massive machine-type communications (mMTC), where a large number of devices transmit sporadic, small packets. In traditional cellular systems, each device is allocated orthogonal resources prior to uplink transmission via a grant mechanism. However, this allocation requires signaling on control channels, which can exceed the data payload size and lead to inefficient resource use. Consequently, grant-free methods [1] that eliminate or reduce control traffic are well suited for these scenarios. Removing coordination introduces non-orthogonality, resulting in the superposition of signals from some or all devices.

A recent family of random access protocols—sometimes called “modern random access”—aims to address and even exploit this phenomenon. The IRSA protocols (Irregular Repetition Slotted ALOHA) [2,3,4] use Successive Interference Cancellation (SIC) and represent one form of grant-free technique, but they can also operate with any packet transmission scheme. They are related to “Unsourced Random Access” [7]. Adapting these methods for grant-free mMTC in cellular networks is therefore of prime interest and the main objective of this position. 

The position is in the framework of dAIEDGE—A network of excellence for distributed, trustworthy, efficient and scalable AI at the Edge—funded by the European Union.

The vision of the dAIEDGE Network of Excellence is to strengthen and support the development of the dynamic European edge AI ecosystem under the umbrella of the European AI Lighthouse and to sustain the advanced research and innovation of distributed AI at the edge as essential digital, enabling, and emerging technology in an extensive range of industrial sectors.

The project is part of the AgroecologIcaL decision making and Optimization with REinforcement learning (Agrilore) project from ANR-TSIA 2025 initiative. This project brings together an interdisciplinary board of researchers from INRIA, CIRAD, and INRAE.

Agroecological intensification is a key response to the current challenges of food security and climate
change \cite{vikas2024agroecological}. Among the agroecological levers, crop diversification, especially intercropping (i.e. growing least two different crop species in the same field), offers major agronomic potential. However, their implementation is still based on limited knowledge. As the production of experimental references is cumbersome and costly, process-based (mechanistic) modeling has emerged as an effective alternative. However, many standard Process-Based crop Models (PBM) including STICS or DSSAT were initially built for monocultures, and while extensions exist, they still struggle to represent all the complex interactions inherent in intercropping systems. In parallel, the crop modeling community is increasingly focusing on issues of model modularity and interoperability, as illustrated by the Crop2ML framework devloped as part of Agricultural Model Exchange Initiative (AMEI).

(18) Postdoctoral Fellowship Position

Postdoc Fellowship Position summary/title: Post-Doctoral Research Visit F/M Postdoc in numerical optimization for statistical machine learning

Deadline : 2026-01-19

View details & Apply

The aim of this project is to develop computationally efficient and robust numerical schemes for simulating multi-material interactions, including interfaces between fluids, gases, and solids. The project uses a monolithic hyperbolic model of partial differential equations to describe all materials within the same Eulerian framework. This allows simulations to be performed across a wide range of flow regimes — from compressible to nearly incompressible — without changing the solver. In addition to interface treatment, the model’s multi-scale nature poses numerical challenges, particularly when stiff materials such as copper or fluids under high pressure are considered. This requires so-called low Mach number schemes, which remain stable when large time steps are taken during deformation or transport of the interfaces while eliminating fast, negligible processes. This necessitates the use of semi-implicit time integrators, which yield non-linear implicit systems. To avoid computational overheads, suitable linearisations are considered that are compatible with the interface conditions, which are treated as immersed boundaries.

This 2-year postdoctoral position is funded by the prestigious Programme Inria Quadrant (PIQ) for the project DynaNova, which aims to advance our understanding of conformational dynamics and allosteric communication in macromolecular complexes. The successful candidate will develop novel graph neural network (GNN) architectures to learn dynamic information from molecular dynamics (MD) simulations of protein-protein and protein-nucleic-acid complexes.

You will join the Delta team at Inria (Université de Lorraine), working closely with Dr. Yasaman Karami, expert in conformational dynamics, allostery, and deep learning for structural biology. The team is growing and offers a highly interdisciplinary environment that brings together researchers in structural bioinformatics, computational chemistry, biophysics, and machine learning.

We have access to major national HPC facilities (Grid5000, Jean Zay, GENCI allocations), including large-scale GPU resources.