AI Scientist, Mistral AI

• PhD or Master's in CS/AI or an engineering science: Mechanical Engineering, Electrical Engineering, Computational Fluid Dynamics, Structural Mechanics, EDA, Semiconductor Engineering, or a related field
• Strong, hands-on machine learning expertise with a deep understanding of model architectures, training dynamics, and evaluation methodology is core to this role
• You have developed ML methods for simulation or surrogate modelling
• You write clean, readable Python code and are comfortable in Linux/HPC environments
• Fluent English with excellent communication skills, able to explain technical simulation and ML concepts to both engineering and non-technical audiences
• Self-directed, you don't need detailed roadmaps to make progress
• Low-ego, collaborative, and eager to learn at the intersection of simulation and ML
• Demonstrated success through industrial projects, academic work, or personal projects

Desirable

• Mistral is looking for AI Scientists with deep expertise in engineering sciences and machine learning to push the frontier of AI-accelerated simulation
• Within AI4Engineering Science, you will research and train foundational physics models which are substantially more capable than what exists today and can be fine-tuned for downstream applications by both customers and internal teams
• You will work across the full research stack: curating high-fidelity simulation datasets, designing and training novel model architectures, and rigorously evaluating them against real engineering validation standards
• Working closely with the broader research organization, you'll ensure the foundation models you build are general enough to become the backbone of many downstream products, not just a single point solution
• This role builds on a strong, world-class foundation, and the goal is to take it further. You'll work one vertical at a time toward foundation models that genuinely transfer and fine-tune across engineering tasks, with high-quality simulation data pipelines, physics-based evaluation, and uncertainty / out-of-distribution estimation as first-class concerns. There's no inherited playbook for most of what's left to do: you'll help define the architectures, training strategies, and validation standards the team builds on, not just extend an existing one
• Research and train novel foundation models for physics simulation, pushing past today's state of the art in accuracy, generalization, and scale
• Design and run large-scale simulation campaigns using domain-specific solvers to build the high-fidelity datasets foundational physics models need
• Investigate architectures and training strategies (e.g. multi-fidelity training, pretraining objectives, scaling behavior) that let a single foundation model transfer and fine-tune well across diverse engineering tasks
• Rigorously evaluate model coverage, accuracy, and robustness against industry validation standards, and diagnose failure modes arising from data gaps or architecture limitations
• Stay on top of the latest developments in the scientific community and contribute to Mistral's standing at the frontier of AI-for-engineering research