Context Soil organic carbon (SOC) plays a crucial role in the global carbon cycle and serves as a key indicator of soil health and ecosystem sustainability. Accurately estimating soil carbon content is essential for climate modeling, sustainable agriculture, and environmental monitoring. Traditional methods based on field sampling and laboratory analysis provide reliable data but are costly, time-consuming, and spatially limited, making large-scale monitoring challenging. With the growing availability of remote sensing data (e.g., Sentinel, Landsat) and complementary geospatial datasets (e.g., climate variables, topography, land use), machine learning approaches have emerged as powerful tools to predict SOC over large areas. Recent advances in deep learning and multimodal data fusion, combining satellite imagery, spectral information, and in-situ measurements, offer new opportunities to improve prediction accuracy and spatial resolution. In this context, this internship aims to develop and evaluate multimodal learning approaches for soil organic carbon estimation, by effectively integrating heterogeneous data sources to build a robust and scalable predictive framework. Missions The intern will contribute to the development of machine and deep learning models for soil organic carbon estimation using heterogeneous data sources. The main objectives and tasks include: Development of a data pipeline integrating diverse datasets, including satellite imagery (e.g., Sentinel, Landsat), climatic and topographic variables, and in-situ soil measurements. Design and implementation of multimodal learning architectures for SOC prediction. Experimentation with various data fusion strategies and learning techniques to improve model robustness and predictive performance. Evaluation of model performance on georeferenced SOC datasets. Required Qualifications The candidate should meet the following criteria: • Education: Currently enrolled in an engineering school or pursuing a Master’s degree (or equivalent); • Technical skills: Strong knowledge of computer vision and deep learning; • Programming: Excellent command of Python and PyTorch; • Ability to work independently with rigor, initiative, and strong organizational skills; • Familiarity with software development best practices is highly appreciated. Duration: 6 months Location: Multitel, Parc Initialis 2, Rue Pierre et Marie Curie, 7000 Mons, Belgium. Application process: Interested candidates should email their application (single PDF named Lastname_Firstname_InternshipTitle.pdf, including CV, cover letter, and academic transcript), indicating their intended start and end dates and the internship title in the email subject line. Depending on the candidate’s profile and interests, the internship may take either a research-oriented or an industry-oriented focus. Multitel is a research and technological innovation center based in Mons, Belgium, supporting industrial players in the development of cutting-edge technological solutions in Artificial Intelligence, Applied Photonics, Networks and Cybersecurity, IoT and Embedded Systems, and Railway Certification. With its multidisciplinary expertise and commitment to excellence, Multitel actively contributes to strengthening industrial competitiveness and fostering innovation at both regional and international levels. Multitel’s Artificial Intelligence Department has strong expertise in machine learning and data-driven technologies, applied to a wide range of data modalities (images, videos, time series, 3D point clouds, spectral and satellite data, and audio signals) as well as diverse application domains (healthcare, agriculture, defense, security, etc.). The department has expertise covering the entire data value chain, from data collection and preprocessing to model development, deployment, and real-world integration