AI Scientist, Agentic Engineering for CAD/CAE & EDA

CDFAM - Computational Design Symposium; France; Remote; EUR 90,000 - 150,000; Full time

Benefits offered by this job

Healthcare coverage

Parental leave

Retirement plans

Relocation support

Wellness programs

Meal and transportation allowances

Job summary

Mistral is seeking an AI Scientist with deep ML expertise to expand agentic tools across the engineering lifecycle, focusing on CAE, CFD/FEA, and EDA/Semi. You will build pre/post-training data and verifiers for Mistral's LLMs to reason about and execute real engineering tasks.

You will collaborate with domain experts to ground workflows in how engineers work, shaping training signals and benchmarks. This is foundational work with no inherited playbook, offering a chance to define what good

Qualifications

• Fluent English with strong communication skills.
• Deep machine learning expertise, with hands-on ML engineering experience.
• Proven ability to design data and verification pipelines for LLMs.

Responsibilities

• Design pretraining, SFT, and RL data for engineering tasks across CAD, CAE, and EDA.
• Define verifiers and evaluation criteria for correctness and quality.
• Design and improve agent architectures and tool-use strategies.
• Build benchmarks and diagnostics to identify model failures and data gaps.
• Collaborate with CAE/EDA domain experts to target high-value workflows.
• Contribute to pre/post-training research and share methodology.

Skills

Fluent English

Deep ML expertise

LLM development

Python programming

Linux HPC experience

Tools

Python

Linux

Job description

Mistral is seeking an AI Scientist with deep ML expertise to expand agentic tools across the engineering lifecycle, focusing on CAE, CFD/FEA, and EDA/Semi. You will build pre/post-training data and verifiers for Mistral's LLMs to reason about and execute real engineering tasks.

You will collaborate with domain experts to ground workflows in how engineers work, shaping training signals and benchmarks. This is foundational work with no inherited playbook, offering a chance to define what good