Post-Doctoral Research Visit F/M Interfacing Crop Models with Reinforcement Learning (F/H)

Inria, the French national research institute for the digital sciences; France; Sur place; EUR 35 000 - 50 000; Plein temps

Résumé du poste

A leading research institute in France is offering a postdoctoral researcher position focused on developing reinforcement learning methods for crop management. The ideal candidate will possess a PhD with strong mathematical skills and a keen interest in applied research at the intersection of agriculture and AI. Responsibilities include designing decision-making algorithms and developing a robust interface for crop simulations. This role offers opportunities for interdisciplinary collaboration and impactful research in agroecology.

Prestations

7 weeks annual leave

Possibility of teleworking

Social and cultural events

Partial reimbursement of public transport costs

Qualifications

• Strong mathematical foundation required for the position.
• Experience in reinforcement learning and adaptive decision algorithms is necessary.
• Strong interest in interdisciplinary research is a must.

Responsabilités

• Develop a robust interface between crop models and reinforcement learning.
• Design adaptive decision-making algorithms for crop simulations.
• Research applications of AI in agriculture.

Connaissances

Mathematical foundation

Probabilistic graphical models

Software optimization experience

Good communication skills

Formation

PhD in relevant field

Outils

Reinforcement Learning frameworks

Crop Simulation Software

Description du poste

Inria, the French national research institute for the digital sciences

Organisation/Company Inria, the French national research institute for the digital sciences Research Field Computer science Researcher Profile Recognised Researcher (R2) Country France Application Deadline 31 Jan 2026 - 00:00 (UTC) Type of Contract Temporary Job Status Full-time Hours Per Week 38.5 Offer Starting Date 1 Mar 2026 Is the job funded through the EU Research Framework Programme? Not funded by a EU programme Reference Number 2025-09673 Is the Job related to staff position within a Research Infrastructure? No

Offer Description

Supervision: The Postdoctoral researcher will be advised by Odalric-Ambrym Maillard from Inria team-project Scool and Cyrille Midingoyi from CIRAD/PERSYST AIDA Unit.

Place: This position will be primarily held at the research center Inria Lille – Nord Europe, Villeneuve d’Ascq, in the Inria team-project Scool (Sequential, Continual and Online Learning), with strong regular interactions with CIRAD AIDA unit in Montpellier.

Keywords: Multi-armed bandits, Sequential statistics, Societal challenge.

INRIA The postdoctoral researcher will be hosted at Centre Inria de l'Université de Lille, in the Scool team. Scool (Sequential COntinual and Online Learning) is an Inria team-project. It was created on November 1st, 2020 as the follow-up of the team SequeL. In a nutshell, the research topic of Scool is the study of the sequential decision making problem under uncertainty. Most of our activities are related to either bandit problems, or reinforcement learning problems. Through collaborations, we are working on their application in various fields including health, agriculture and ecology, sustainable development. More information, please visit https://team.inria.fr/scool/projects

Odalric-Ambrym Maillard is a permanent researcher at Inria. He has worked for over a decade on advancing the theoretical foundations of reinforcement learning, using a combination of tools from statistics, optimization and control, in order to build more efficient algorithms able to provide decision making in uncertain environments. He was PI of several projects, including ANR-JCJC project BADASS (BAnDits Against non-Stationarity and Structure), Inria Action Exploratoire SR4SG (Sequential Recommendation for Sustainable Gardening) and Inria-Japan Associate team RELIANT (Reliable Bandit strategies). His goal is to push forward key fundamental and applied questions related to the grand-challenge of making reinforcement learning applicable in real-life societal applications.

Context 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). This project aims to overcome these limitations by integrating artificial intelligence (AI) approaches, notably reinforcement learning (RL), to optimize decision‑making under conditions of uncertainty. Building on a proven interdisciplinary collaboration in simpler intercropping contexts, we leverage this dynamic to tackle a major challenge: adapting the RL‑Agro coupling to the emblematic case of intercropping. Beyond its agronomic importance, this research also contributes to the rapidly growing intersection between AI and environmental modeling. Reinforcement learning environments inspired by complex natural systems are gaining traction in the machine learning community as challenging, high‑dimensional testbeds. Notably, the recently developed WOFOSTGym simulator \cite{solow2025wofostgym}, bridging crop modeling and RL, received the Outstanding Paper Award at the 2024 Reinforcement Learning Conference, highlighting the community’s enthusiasm for scientifically grounded RL environments. Developing a generic Gym‑PBM, therefore, not only supports sustainable agriculture but also provides the RL community with a novel, physically grounded benchmark characterized by temporal dependencies, partial observability, and uncertainty—features often missing in standard RL benchmarks. The system’s modular design will enable broader methodological experimentation and open the door for AI researchers to engage with real‑world, sustainability‑driven challenges.

By positioning the project at the interface between agronomy, AI, and applied mathematics, this research contributes to the emergence of a new interdisciplinary field where model‑based reasoning and data‑driven learning co‑evolve. This synergy is expected to foster new collaborations between RL researchers and agricultural scientists, accelerating innovation in both domains.

Objectives The goal of the postdoc project is to develop a robust and flexible interface between crop models and reinforcement learning (RL) to enable decision‑making algorithms to interact with crop simulations. This requires bridging a significant conceptual and technical gap between monolithic crop models, in which user‑defined management actions are treated as predefined input variables, and RL, which relies on state‑action‑reward loops to enable adaptive decision‑making under uncertainty.

The core challenge, therefore, is to design a modular and generalizable methodology that embeds STICS within a standard RL framework. This will pave the way for adaptive, data‑driven decision‑making in agronomy and open new opportunities to optimize crop management strategies in complex, uncertain environments.

Methodology

The methodological pathway follows a generic‑to‑specific progression: defining a model‑agnostic formalism to couple black‑box PBM with reinforcement learning (RL), then implementing this abstraction in a reusable software interface, and finally instantiate and evaluate it on STICS or other PBM.

• O1 – Define a model‑agnostic formalism for RL–PBM coupling. The first step is to develop an abstract representation of how an RL agent can interact with a PBM. It will focus on the formalization of how to access state variables or other indicators from crop model run, how to change recommendations or management decisions (actions) through controllable input levers (e.g. sowing date, fertilization, irrigation, cultivar choice), and how to define reward signals that encode agronomic and environmental objectives (yield, risk, resource use, emissions, etc.). This formalism will be expressed as a generic state–action–reward–transition schema, compatible with standard RL frameworks, but enriched with agronomic structure (time step, crop management, etc.). This model‑agnostic abstraction will serve as a conceptual template for any mechanistic crop model.
• O2 – Implement a generic RL–crop interface library. Building on the abstract formalism defined in O1, the second objective is to implement a generic software interface that operationalizes this coupling, in the form of an OpenAI Gymnasium–compatible environment layer (a “Gym‑Agro” or ”Gym‑PBM abstraction). This layer will expose a standardized API (reset, step, observe, reward, done) to RL agents, manage simulation calls and time‑stepping, allow the mapping between model‑specific I/O and the abstract state–action–reward schema to be specified declaratively, etc. The result will be a model‑independent environment layer into which different crop models can be plugged without changing the core RL code, simply by providing a suitable adapter specification.
• O3 – Instantiate and evaluate the formalism on STICS. The third objective is to specialize and validate the generic methodology on STICS. Using the formalism and the “Gym‑Agro” interface from O1–O2, we will develop a STICS‑specific adapter that maps STICS input variables and management options to actions, extracts relevant biophysical and management indicators as states, defines appropriate reward functions in line with agronomic objectives. We will demonstrate the flexibility and scalability of the approach for diverse cropping systems and objectives.

Excellent writing and presentation skills

Excellent organisation and communication skills due to interdisciplinary context.

Specific Requirements

Desired Postdoctoral Profile and Scientific Challenges

The postdoctoral position requires both a strong mathematical foundation and a deep interest in applied interdisciplinary research. The coupling of RL with mechanistic crop models entails challenges beyond standard decision‑making frameworks: calibration, inference, and optimization must operate over high‑dimensional, continuous, and structured variable spaces.

In this context, expertise in probabilistic graphical models and non‑parametric inference methods will be crucial to learn structural dependencies and perform efficient inference under data scarcity. The project also opens perspectives for applying and extending continuous Bayesian networks and bandit‑based experimental design approaches to agricultural systems.

Because of the computational demands of crop simulations, software optimization and scalability are also central. The candidate should ideally have experience implementing large‑scale inference or optimization methods, and a strong interest in reinforcement learning and adaptive decision algorithms. This position thus offers a rare opportunity to combine advanced mathematical modeling with real‑world impact in agroecology.

Languages FRENCH Level Basic

Languages ENGLISH Level Good

Additional Information

• Partial reimbursement of public transport costs
• Leave: 7 weeks of annual leave + 10 extra days off due to RTT (statutory reduction in working hours) + possibility of exceptional leave (sick children, moving home, etc.)
• Possibility of teleworking and flexible organization of working hours
• Professional equipment available (videoconferencing, loan of computer equipment, etc.)
• Social, cultural and sports events and activities