AI Scientist, Agentic CAD (AI4Engineering)
Paris • Full-time
Posted 8d ago
Job Location
Paris
Tech Stack
Remote Work Policy
On-site
Employment Type
Full-time
Categories
AI Agent Engineer
About the job
Mistral is seeking an AI Scientist with deep machine learning expertise and hands-on CAD experience to advance the capabilities of our agentic tools for computer-aided design. The primary mission is to train Mistral's LLMs to generate, analyze, and modify CAD designs with the proficiency of an expert design engineer. This involves creating pre- and post-training data for LLMs to execute real CAD tasks, such as text-to-CAD and image-to-CAD generation, geometric analysis, and design modification. The role offers a unique opportunity to directly influence models by designing the data and verifiers that are incorporated into them. Additionally, you will contribute to shaping agent architectures and harnesses that enable reliable, multi-step CAD workflows through API integrations and computer use.
Responsibilities
- Design pretraining, SFT, and RL data for CAD tasks including text-to-CAD and image-to-CAD generation, geometric analysis, and design modification.
- Define verifiers and evaluation criteria for CAD tasks to ensure correctness and quality.
- Scale data pipelines for producing agentic CAD training traces.
- Design and improve agent architectures and harnesses for planning, tool calling, error recovery, and multi-step workflows.
- Build and maintain CAD evaluation benchmarks and diagnostic tooling to identify and trace model failures.
- Collaborate with CAD domain experts and other teams to identify high-value CAD workflows.
- Contribute to broader pre- and post-training research within the science organization.
Requirements
- Deep, hands-on machine learning expertise, particularly in LLM development (SFT, RL, fine-tuning, GRPO) demonstrated through publications or shipped systems.
- Demonstrated hands-on CAD experience, including designing, scripting, or automating CAD workflows in tools like SolidWorks, CATIA, Siemens NX, FreeCAD, or CADQuery, ideally with CAD-kernel grounding.
- Fluent English with excellent communication skills for explaining technical concepts.
- Proficiency in writing clean Python code.
- Comfortable working in Linux/HPC environments.
- Self-directed and able to make progress without detailed roadmaps.
- Low-ego, collaborative, and eager to learn at the intersection of CAD and ML.
Benefits
- Comprehensive benefits package designed to support well-being, growth, and work-life balance.
- Healthcare coverage.
- Parental leave.
- Retirement plans.
- Relocation support.
- Wellness programs.
- Meal and transportation allowances.
- Other location-specific perks.
