This is a competitive application. The doctoral schools offer PhD positions for the best ranked students of each master program. MSIAM typically gets 3 positions.
The fellowship, though granted by doctoral schools, actually come from the French Ministry for Education and Research.
In principle the students at Ensimag and UGA IM2AG apply to the MSTII doctoral school.
CNRS offers so-called BDI PhD Fellowships http://www.dgdr.cnrs.fr/drh/emploi-nonperm/formation.htm
There is no call for application. The applicant must register to a doctoral school, contact a laboratory, define his/her project and apply to the BDI fellowship.
The PhD project may (but does not have to) include some industrial partner.
A small number of PhD positions are offered by Inria https://www.inria.fr/en/centre/sophia/overview/offers/phd-positions/phd-positions
https://www.inria.fr/recherches/jeunes-chercheurs/etre-doctorant/mode-d-emploi
Inria offers a closed list of PhD positions. If you want to define your own PhD project in association with some Inria researcher to make it eligible (i.e. in the closed list), it probably has to be prepared one year before the beginning of the PhD.
Quite a lot of other public research institutes have PhD positions, among which CEA, INSERM, INRA and IRSTEA.
A CIFRE PhD is a tripartite agreement signed by the PhD student, an industrial and a research laboratory.
All the CIFRE PhDs are managed by the ANRS
Research achieved during a CIFRE PhD will be applied research on a subject proposed by an industrial. The approach will thus be clearly a scientific approach as in any thesis but concrete fallouts are expected by the industrial partner.
Works could result in concrete applications, publications in journals or conferences and/or patent registration.
The 71 universities in France each have their own calls to projects for research. Groups of researchers from different labs get support for specific projects, which often include 1 or 2 PhD fellowships.
This used to be the case of Labex (groups of laboratories).
The Rhône-Alpes Region allocates doctoral research grants every year. Its 8 Academic Research Communities (Communautés Académiques de Recherche [ARC]) also finance doctoral dissertations in subjects corresponding to their specialties. We are typically involved in ARC 6 Theme 2: Dispositifs, systèmes, calcul et logiciels (http://lig-membres.imag.fr/ledru/ARC6/FicheARC6v3.5.pdf unfortunately in French).
Warning, the grant is requested by the advisor (AAP=“Project Call” quite early Feb or March -> ask your MSc thesis advisor)
Once hired into a PhD position, you can subscribe to different variants of contract.
One of them makes it possible to receive some training to be a teacher, and to give courses (or lab works or tutorials) at university and receive some additional wages for that.
In our jargon this contract is called DCE or RES (see “Research and Higher Education”) or B contract (!!!) or monitor.
Ce sujet de thèse en statistiques mathématiques, financé, devrait probablement être mis au concours de l’école doctorale pour un démarrage en septembre 2026. Il s’adresse aux étudiants de niveau M2 ayant un solide parcours en mathématiques et souhaitant s’intéresser à la théorie des statistiques.
Recent advances in neuroscience and machine learning have revitalized interest in models of noisy cognition, which depart from the classical view that decisions result from stable preferences. Instead, these models attribute behavioral variability to systematic limitations in cognitive processing. This framework has gained particular traction in the study of decision-making under risk and uncertainty, offering explanations for well-documented behavioral regularities, such as risk aversion, loss aversion, and probability weighting. In this project, we aim to extend the cognitively grounded models to simple strategic interactions, including social dilemmas, coordination problems, and bargaining/trust games, and then to design and implement behavioral experiments to empirically test the predictive validity of these models against traditional utility-based models.
Recent advances in neuroscience and machine learning have revitalized interest in models of noisy cognition, which depart from the classical view that decisions result from stable preferences. Instead, these models attribute behavioral variability to systematic limitations in cognitive processing. This framework has gained particular traction in the study of decision-making under risk and uncertainty, offering explanations for well-documented behavioral regularities, such as risk aversion, loss aversion, and probability weighting. In this project, we aim to extend the cognitively grounded models to simple strategic interactions, including social dilemmas, coordination problems, and bargaining/trust games, and then to design and implement behavioral experiments to empirically test the predictive validity of these models against traditional utility-based models.
In standard models, agents are assumed to have an exogenously given correct model of the underlying uncertainty. In practice, the “correct model” is rarely known and has to be inferred from data. Both humans and AI-decisions makers may experience surprises and have to take into account “unknown unknowns”. This project studies the way in which humans and machines derive the relevant model of uncertainty from available data in view of possible unforeseen contingencies. The goals of the project are three-fold: derive and study an axiomatic model of decision making under uncertainty, applicable to both human and machine-learning. Use the model to generate behavioral and normative predictions, as well as to study learning and the value of information acquisition in such environments. Design an experiment to test the theoretical predictions using human subjects and AI-algorithms.
In the context of research actions in the field of the Security of Artificial Intelligence, we propose the following PhD subject : Security of decentralized deep learning : integrity and confidentiality of embedded deep neural network models
Definition of the different threats models for embedded and decentralized training (focus on Federated Learning)
contact : pierre-alain.moellic@cea.fr
LORIA, Paris colorisation automatique d’images et de vidéos à l’aide de réseaux de neurones et méthodes variationnelles
LJK, Grenoble DISTRIBUTIONALLY ROBUST SHAPE OPTIMIZATION