Missions

The postdoctoral researcher's main task will be to develop large-scale deep learning approaches to analyze multi-source time series of satellite images (Sentinel-1/2, Pléiades, Pléiades-NEO, SPOT 6/7, etc.) and other complementary spatial or environmental modalities. The objective will be to explore new foundational representations for understanding surface dynamics and underlying processes at different spatial and temporal scales. This work is part of an interdisciplinary framework combining artificial intelligence, geomatics, and environmental sciences, and contributes to the development of generic approaches for observing environmental dynamics based on massive Earth observation data.

Activités

- Design and implement a processing chain that uses multi-modal and multi-temporal satellite data with high spatial resolution.
- Develop and adapt large-scale learning models (“giga-models” or foundation models) for characterizing and monitoring environmental dynamics.
- Integrate and harmonize time series from heterogeneous sensors (satellites, in situ, participatory, etc.) to improve understanding of the spatial and temporal evolution of the environments observed.
- Contribute to the scientific dissemination of results through international A-level publications and actions to open up the data and models developed.

Compétences

The candidate will hold a PhD in geomatics or geography. He/she will have advanced expertise in image processing (machine/deep learning), spatial analysis, and statistics. Advanced skills in image time series processing methods are required, as well as in geomatics programming languages such as Python and R, and new developments in deep learning (e.g., foundation models, self-supervised learning) with experience in cloud computing environments.
Accordingly, a working proficiency in both French (C1) and English (B2) sufficient for smooth communication is essential. The position requires a high level of autonomy in data processing and strong scientific writing skills in English (B2).

Contexte de travail

The position involves geographical sciences, remote sensing, and scientific computing. The research works are part of the M2-BDA project (Multi-Modal Big Data Analytics: Knowledge Distillation and Deep Learning for Actionable Insights), funded by the French National Research Agency (ANR), which aims to develop innovative methods for processing and analyzing heterogeneous Earth Observation data — including in situ, airborne, satellite, and participatory sources.
The project focuses on designing advanced learning approaches capable of handling large-scale, multi-source, and multi-scale data streams. It addresses several major scientific challenges: (1) Managing multi-modal, temporal, and multi-resolution datasets ; (2) Developing giga-models and self-learning strategies to overcome the lack of annotated data ; (3) Applying knowledge distillation techniques to simplify complex deep learning architectures, and (4) Enhancing model explainability to better understand and interpret deep network behavior.
The postdoctoral researcher will be based at the Image, City, Environment Laboratory (LIVE, UMR 7362 CNRS / University of Strasbourg), within the Faculty of Geography and Spatial Planning. The host laboratory is ideally located on the University's main campus, close to the city center and easily accessible by public transportation. It is surrounded by all essential amenities, including restaurants and housing options.
The project brings together experts from computer science (Laboratoire d'Informatique Paris Descartes - LIPADE, Institut de Recherche en Informatique, Mathématiques, Automatique et Signal - IRIMAS), environmental geography (LIVE), and geosciences (Institut Terre et Environnement de Strasbourg - ITES), in collaboration with the Data Terra Research Infrastructure (RI). This infrastructure supports the FAIRification and dissemination of data and results, as well as the operationalization of processing workflows, through the A2S Data and Services Center (EOST/Unistra), linked to the University of Strasbourg Data Center. As part of a multidisciplinary team with multiple partners from a variety of scientific backgrounds, the candidate must be able to collaborate effectively and act as a bridge between the different project teams. Fieldwork experience or a strong interest in field activities would be an asset.

Contraintes et risques

No associated risks.