> Markdown version of [/jobs/ext/4497637-job-offer](https://www.wearedevelopers.com/jobs/ext/4497637-job-offer). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Job offer

- **Company:** BRGM
- **Location:** Orléans, France (Remote available)
- **Contract:** Temporary contract
- **Skills:** Artificial Intelligence, Computer Vision, Computer Programming, Image Analysis, Data Fusion, Python (Programming Language), Machine Learning, Tensorflow, Scientific Computating, Software Engineering, Computational Statistics, Multimodal Learning, Pytorch, Transfer Learning, Uncertainty Quantification (UQ), Deep Learning, Model Validation, Feature Extraction, Lidar
- **Published:** November 2, 2026
- **Apply:** https://brgm-recrute.talent-soft.com/job/emploi-post-doct-apprentissage-automatique-pour-la-modelisation-predictive-du-comportement-de-fragmentation_4115.aspx?LCID=2057

## About the Role

You hold a PhD in Machine Learning, Artificial Intelligence, Data Science, Applied Mathematics, Computational Statistics, Computer Vision, Scientific Computing, or a closely related discipline. You are interested in applying advanced machine learning methods to challenging real-world scientific problems.
Ideally, you should have:
-Strong expertise in machine learning and deep learning.
-Experience with multimodal learning, uncertainty quantification, probabilistic modeling or related approaches.
-Knowledge of scientific machine learning, physics-informed learning, hybrid modeling.
-Experience with image analysis, computer vision, geometric data, point clouds, graphs, or other structured data representations.
-Strong programming skills in Python and modern ML frameworks such as PyTorch, TensorFlow.
-Experience with reproducible research, model evaluation, benchmarking, and scientific software development.
-Strong communication and collaboration skills in multidisciplinary research environments.
-Experience in mineral processing would be a plus., PhD or equivalent

## Description

As part of the PEPR Sous-sol, bien commun program, the InnovTech project federates eight French research institutions around the sustainable processing of mineral resources, with an explicit focus on artificial intelligence and machine learning. This postdoctoral position is part of Sub-Task ST3.1.1: Closing the gap between elementary ore fragmentation properties and the performances of comminution technologies (Leader: INP/LGC - Participants: BRGM, LGF/MSE). Comminution accounts for approximately 50% of the energy consumption of world mineral processing plants. ST3.1.1 proposes a bottom-up methodology: starting from particle-scale fragmentation data generated by dedicated experimental tools and DEM simulations, coupled machine learning and phenomenological models will bridge elementary fragmentation properties and the performance of comminution machines. The ML problem is the following. Three complementary data sources characterize each mineral particle: 3D micro-CT volumes (crack
networks, porosity, mineralogy), single-particle breakage test results (fracture energy, breakage probability, fragment size distributions), and DEM simulation outputs (fracture mechanisms, energy dissipation)., As a postdoctoral researcher within InnovTech ST3.1.1, you will contribute to the development of next-generation machine learning methods for scientific and industrial data.
The project focuses on understanding and predicting the fragmentation behavior of mineral particles from heterogeneous experimental observations.
Key research directions include:
-Multimodal learning and data fusion: Develop machine learning approaches capable of integrating heterogeneous scientific and industrial data sources, including imaging, material characterization, mechanical measurements, and fragmentation experiments.
-Representation learning for complex material systems: Design compact, interpretable, and transferable representations that characterize particle properties and fragmentation behavior.
-Probabilistic machine learning and uncertainty quantification: Develop models that predict distributions and confidence estimates rather than single deterministic outputs.
-Multi-scale predictive modeling: Develop models capable of linking observations collected at different spatial and temporal scales, from material and particle characterization to fragmentation behavior and process-level performance.
-Data-efficient machine learning: Explore methodologies suited to limited, heterogeneous, and partially observed datasets, including transfer learning, self-supervised learning, surrogate modeling, active learning, and uncertainty-aware decision-making approaches., This position is to be filled within the DCGS/GDS unit of the BRGM at the technical center of Orléans (45) You will work closely with a multidisciplinary team of researchers from data science, applied mathematics, process engineering, and materials science. The position includes leading research activities, publishing in high-impact journals, contributing to project deliverables, and presenting results at international conferences.