We are evolving from a classic Data Science function toward an AI-native team: embedded alongside Product, R&D, GTM and Design, we frame the right problems early, prototype fast, and build trusted, reusable AI systems that help Criteo ship better products, faster, harnessing Product Intelligence. Agent evaluation & guardrails - help build frameworks that evaluate multi-turn agents (LLM-driven personas talk to a real agent, LLM judges score relevance, coherence, helpfulness, safety and accuracy), so agentic products ship faster and more safely.