Evaluation Engineer, Mistral AI

• You are fluent in English
• 3+ years of experience in ML evaluation, benchmarking for LLM or agentic systems
• You have proven experience in AI or machine learning product implementation with APIs, back-end
• You have deep understanding of concepts and algorithms underlying machine learning and LLMs
• You have strong technical coding skills in Python
• You hold strong communication skills with an ability to explain complex technical concepts in simple terms with technical and non-technical audiences

Desirable

• The Applied AI team is Mistral's customer-facing technical organization
• We work directly with enterprise clients from pre-sales through implementation to deploy cutting-edge AI solutions that deliver measurable business impact
• Our team combines deep ML expertise with strong customer engagement skills, operating like startup CTOs who own end-to-end project execution
• However, the AI graveyard is full of great ideas nobody could measure or prototypes that never made it to production
• As a first Evaluation Engineer, you'll design the methodology, build the infrastructure, and define what "ready for production" means across verticals and use cases
• You will design and implement evaluation systems that help our customers understand model performance across their specific use cases, build robust evaluation infrastructure, and work closely with both research and customer-facing teams
• Research builds evals for frontier capabilities but customers don't care about MMLU scores
• We need in Applied AI evals and frameworks for customer reality domain-specific, risk-aware, production-grade
• The kind that tell you whether your medical summarization model will hallucinate drug interactions, or whether your legal assistant will invent case citations
• This role sits at the intersection of research, engineering, and solutions, you will play a critical cross role in measuring, understanding, and improving the capabilities of our models for our enterprise customers
• Design and implement comprehensive evaluation frameworks to measure LLM capabilities across diverse customer use cases, including text generation, reasoning, code, and domain-specific applications
• Build scalable evaluation infrastructure and pipelines that enable rapid, reproducible assessment of model performance
• Develop novel evaluation methodologies to assess emerging capabilities or verticalized use cases (cybersecurity, finance, healthcare, etc.) and enable the Solutions (Deployment Strategist and Applied AI) on these topics
• Create custom evaluation suites tailored to enterprise customers' specific needs, working closely with them to understand their requirements and success criteria
• Collaborate with research teams to translate evaluation insights into model improvements and training decisions
• Partner with product teams to continuously improve our evaluation tooling based on customer feedback
• How We Work in Applied AI:
• We care about people and outputs
• What matters is what you ship, not the time you spend on it
• Bureaucracy is where urgency goes to vanish. You talk to whoever you need to talk to. The best idea wins, whether it comes from a principal engineer or someone in their first week
• Always ask why. The best solutions come from deep understanding, not from copying what worked before
• We say what we mean. Feedback is direct, timely, and given because we care
• No politics. Low ego, high standards
• We embrace an unstructured environment and find joy in it