AI Scientist - Agentic Engineering

About Mistral Mistral provides full-stack AI solutions: from frontier models to developer tools, applications, and compute. We partner with enterprises tackling the hardest problems—across high-stakes industries like finance, manufacturing, defense, healthcare, and the public sector—co-creating customized AI systems that they can run on their terms. We are a dynamic, collaborative team passionate about AI and its potential to transform society. Our diverse workforce thrives in competitive environments and is committed to driving innovation. Our teams are distributed between Europe, North America, Asia and the Middle East. We are creative, low-ego and team-spirited.

Mistral is looking for AI Scientists with deep ML expertise and hands-on engineering experience to expand what our agentic tools can do across the engineering lifecycle — CAE (CFD, FEA, etc.) and EDA/Semi. Working within the AI4Engineering Science team, your core work is building the pre and post-training data for Mistral's LLMs to reason about and execute real engineering tasks. Because Mistral trains its own frontier LLMs, the data and verifiers you design ship directly into models you can hold, a rare position, and the core of the job. Alongside this, you'll help shape the agent architectures and harness that let these models operate reliably inside multi-step engineering workflows, not just answer isolated questions. You'll work closely with domain experts across CAE and EDA or other domains to ground this work in how engineers actually work, and with the broader research team to translate that domain grounding into training signal and evaluation benchmarks that measure genuine task competence. This is early-stage work, and that's the point: you'd be joining at the foundation, shaping the data, verifiers, and agent scaffolding that decide whether these systems become reliable or stay demo-grade. There's no inherited playbook, you'll help define what good looks like, and your work will set the direction the team builds on rather than extend an existing one.

Design pretraining, SFT, and RL data for engineering tasks across CAD, CAE, and semiconductor/EDA Define verifiers and evaluation criteria that capture what "correct" and "high-quality" actually mean for each engineering task, beyond surface-level plausibility Design and improve agent architectures and harnesses: how models plan, call tools, recover from errors, and chain steps together across long-horizon engineering workflows Build evaluation benchmarks and diagnostic tooling to identify where models fail on engineering tasks, and trace those failures back to gaps in data, reward design, or agent scaffolding Collaborate with domain experts across CAE and EDA or other domains (and the science and solutions teams more broadly) to identify which engineering workflows are highest-value to target next Contribute to Mistral's broader pre and post-training research, sharing findings and methodology across the science organization What we're looking for Fluent English with excellent communication skills, able to explain technical ML and engineering concepts to both engineering and non-technical audiences Deep, hands-on machine learning expertise, particularly LLM development Demonstrated experience running, debugging, and validating real engineering workflows in at least one of CAD, CAE, semiconductor simulation, or EDA You write clean, readable Python code and are comfortable in Linux/HPC environments Self-directed, you don't need detailed roadmaps to make progress Low-ego, collaborative, and eager to learn at the intersection of engineering and ML It would be great if you Have experience building or fine-tuning agentic systems (tool use, multi-step planning, agent orchestration frameworks) Have experience with reward modeling, RLHF/RLAIF/RLVR, or preference-based training Have industrial or academic experience with CAE or EDA tools (e.g. SolidWorks, CATIA, Fluent, Abaqus, LS-DYNA,