AI Scientist, Agentic CAD (AI4Engineering)

Location: Hybrid, with positions in Paris, Palo Alto, Amsterdam, Linz, Munich, London, or Berlin

Company

Mistral AI provides full-stack AI solutions, including frontier models, developer tools, applications, and compute, for high-stakes industries such as finance, manufacturing, defense, healthcare, and the public sector.

What you will do

• Design pretraining, SFT, and RL data for text-to-CAD, image-to-CAD, geometric analysis, CAD quality assurance, and design modification.
• Define CAD verifiers, evaluation criteria, benchmarks, and diagnostic tooling for parametric CAD, drawing-to-3D, and industrial workflows.
• Scale pipelines that produce agentic CAD training traces.
• Design agent architectures and harnesses for planning, CAD tool calls through APIs or computer use, error recovery, and long-horizon workflows.
• Collaborate with CAD experts and CAE, EDA, science, and solutions teams to identify high-value workflows.
• Contribute to broader pre- and post-training research and share methodology across the science organization.

Requirements

• Deep, hands-on machine learning expertise, particularly LLM development, including demonstrated experience building LLM training pipelines with SFT, RL, fine-tuning, or GRPO through publications or shipped systems.
• Hands-on CAD experience designing, scripting, or automating real workflows with tools such as SolidWorks, CATIA, Siemens NX, FreeCAD, or CADQuery.
• Fluent English with excellent communication skills for explaining technical ML and CAD concepts to engineering and non-technical audiences.
• Clean, readable Python code and comfort working in Linux/HPC environments.
• Self-directed approach and ability to make progress without detailed roadmaps.
• Low-ego, collaborative working style and willingness to learn at the intersection of CAD and ML.

Nice to have

• CAD-kernel grounding with OpenCASCADE or build123d.
• Experience building or fine-tuning agentic systems, including tool use, multi-step planning, or agent orchestration frameworks.
• Experience with reward modeling, RLHF, RLAIF, RLVR, or preference-based training.
• First-author publications or open-source benchmark work in agentic CAD or CAD-ML.
• Contributions to a large open-source or industry codebase.
• Publications in engineering and ML venues such as NeurIPS, ICLR, ICCV, SIGGRAPH, JFM, or AIAA.

Culture & Benefits

• Full-time hybrid work across the listed locations.
• Collaborative, low-ego, team-oriented environment with teams distributed across Europe, North America, Asia, and the Middle East.
• Benefits may include healthcare, parental leave, retirement plans, relocation support, wellness programs, and meal and transportation allowances, varying by country.