Postdoctoral Researcher (M/F) – Development of an agent-based AI environment for distributed control of power grids

Job Description

Missions

The pace of advancement in various artificial intelligence tools has accelerated significantly in recent years, both for conversational agents (LLMs) and for task-oriented agents. These advances are challenging the added value of many professions, while simultaneously presenting tremendous opportunities. In the field of electrical engineering research, advancements are already evident in the assimilation of state-of-the-art knowledge through exhaustive literature reviews, the rapid development of complex and high-performance simulation codes, and assistance with the drafting of reports and scientific articles. As these tools become more powerful, the role of human researchers in the development of new knowledge is gradually being called into question.
The objective of this research project is therefore to explore what can be achieved using existing agent-based and conversational tools, and to coordinate them to create an environment that takes into account the various phases of research work. Beyond the tasks performed, the motivation behind this postdoctoral research project is thus to identify which components of academic work will offer human added value in the future. Some tasks will be delegated, in whole or in part, to agents. Others will continue to involve interaction between the human researcher and their tool. Finally, this new division of labor will undoubtedly lead to the development of a new role for human intelligence in research work, which this project aims to define.

Activity

The postdoctoral researcher will be responsible for developing an agent-based AI environment. The purpose of this environment will be to assist researchers (regardless of their experience level—whether they are doctoral students, postdoctoral researchers, or tenured faculty) with the core tasks of scientific work: literature review, formulation of a scientific hypothesis, development of simulation software, conducting experimental designs, and writing reports or articles. This environment will be applied to research problems drawn from the state-of-the-art in power grid management. Problems for which a solution already exists will be considered first in order to establish an existing baseline against which the results proposed by AI can be compared. Subsequently, open problems may also be investigated. These will primarily involve high-dimensional stochastic optimal control problems that admit efficient decompositions.
Since research topics are, by definition, aimed at generating knowledge controlled by their developers, this project will seek to use local computing resources rather than remote computing servers. This will, on the one hand, allow for the precise specification of the computing resources needed to run the models under investigation, as servers accessible via the cloud do not allow for an accurate measurement of the computing costs incurred. Furthermore, the process of creating reproducible knowledge for a research team requires the use of open-weight models, executed locally on specific machines. This approach also ensures that confidential datasets or original ideas—which cannot be incorporated into proprietary model training datasets—are not disclosed. The development of this environment will rely on existing tools. However, the construction of the environment itself is an integral part of the research approach for this postdoctoral project. Indeed, the organization of the various agents used, as well as their interactions with one another and with the researcher, must be studied in order to formulate recommendations for achieving relevant results. Since the available open-weight models cover a range of trade-offs between complexity, required memory size, performance, and execution time, this research must investigate the optimal combinations that yield the best results. The necessary computational resources, as well as the reflection time allotted, will be the main criteria for comparing different organizational approaches: the number of agents running in parallel, the types of agents used, autonomous reflection time, etc. Documenting the proposed environment will also be crucial. The goal is to build a tool that can be used by other researchers on the team, initially under the supervision of the postdoctoral fellow and later independently. The researcher hired is expected to be actively involved in the team's activities: by taking an interest in the doctoral students' research topics and advising them on how they might use AI tools, by participating in the organization of the team and its computing resources, and/or by assisting with the search for funding. These interactions will be all the more crucial given that the research project aims to develop a tool that can be used by other researchers. Participation in teaching activities within a teaching and research department at ENS Rennes is possible if the postdoctoral fellow so desires.

Your Profil

Skills

The successful candidate must hold a Ph.D. in electrical engineering, signal processing, or computer science. This position requires multidisciplinary experience. Knowledge of applied computer science, particularly in machine learning, will be essential. The candidate's prior experience using AI tools for research purposes will be a key factor in the selection process. In addition, disciplinary knowledge in the field of application—electrical engineering, state estimation, and optimal control—will be valuable. Additional qualifications, such as experience in training and supervising master's students, and the ability to work as part of a team interacting with several doctoral students and faculty members, will also be valuable assets.

Your Work Environment

The successful candidate will be based at ENS Rennes, located on the Ker Lann campus (Bruz). ENS Rennes is home to the SyEnsCES (Sustainable Energy Systems and Electrical Transducers) team at the IETR laboratory (UMR CNRS 6164). This is a joint research unit specializing in electronics and digital technologies. The SyEnsCES team conducts research on sustainable energy systems and power grid management. The proliferation of renewable energy assets and flexible consumption patterns requires the design of distributed control algorithms capable of being deployed across a very large number of participants. Their ability to make optimal decisions when multiple variables are subject to randomness is therefore crucial. Once validated through simulation, these new decentralized algorithms must be experimentally confirmed through implementation on the team's experimental platform, which enables the emulation of energy flows within a network. Like the researchers at the IETR at the École Normale Supérieure de Rennes (ENS Rennes), the candidate will be part of the Mechatronics Department. This academic department offers programs ranging from the L3 to the M2 levels, focused on careers in teaching and research in industrial sciences. The candidate will be surrounded by fellow doctoral students, instructors, professors, and staff. The professional environment places great emphasis on training and the professional development of all members of the department.

Constraints and risks

ZRR Authorization Required

Compensation and benefits

Compensation

Between €3,041.58 and €3,467.33

Annual leave and RTT

44 jours

Remote Working practice and compensation

Pratique et indemnisation du TT

Transport

Prise en charge à 75% du coût et forfait mobilité durable jusqu’à 300€

About the offer

About the CNRS

The CNRS is a major player in fundamental research on a global scale. The CNRS is the only French organization active in all scientific fields. Its unique position as a multi-specialist allows it to bring together different disciplines to address the most important challenges of the contemporary world, in connection with the actors of change.

CNRS

The research professions