• Design, build, and operate AI agents in production (multi-agent architecture: pre-processing, deterministic routing, LLM-as-judge, SOP-driven agents).

• Ship fast and often: daily iterations driven by real bugs and tickets from production, not by R&D.

• Deploy and operate autonomously: CI/CD, sandbox/production environments, deployment scripts, monitoring (Langfuse, observability for resolver agents).

• Design agent evaluations: evals, LLM-as-judge, fixed-context replay, regression tests : if we can't measure it, we don't ship it.

• Build solid local test environments: reproduce before you fix.

• Integrate internal data to build relevant context accurately and in a controlled way (context engineering).

• Discover needs directly with business teams (CX, RevOps, Data), hand agents over in forward-deployed-engineer mode, and contribute to the architectural vision of the agent platform.

Must-have:

• 5+ years shipping and running production systems, with production-grade code and a real sense of urgency.

• Whole-system thinker: you challenge architecture decisions and weigh the business impact of technical choices.

• Comfortable with ambiguity: you turn a fuzzy problem into a shipped solution without a pre-written spec.

• DevOps-fluent: own deployment end-to-end, infra-as-code, debug network environments (VPN, proxies).

• Agents & LLMs : Hands-on experience shipping agents to production (frameworks such as ADK, LangChain, etc.), applied prompt engineering, evaluation.

• Hands-on experience shipping agents to production (ADK, LangChain, or similar), with applied prompt engineering and evaluation.

• You build with AI daily (Claude Code, Cursor, or equivalent) — not just building agents.

• Fluent in French and English.




Nice to have:




• ADK v2 (Google Agent Development Kit) and the A2A protocol.

• Langfuse or an equivalent LLM observability tool.

• LLM cost optimization at scale (caching, routing, model selection).

• Tech lead / CTO background, or forward-deployed-engineer experience.

• A rare playground: making the operations of a profitable scale-up AI native, with agents running in production and visible impact every week.

• An AI-native environment: access to best-in-class tools (Claude, internal agentic tooling, dedicated observability) and an AI-assisted engineering culture.

• Real autonomy: you define problems together with business teams and carry your solutions all the way to production.

• You're looking for a research-oriented data science role : we need a builder, not an R&D profile.

• You want a pure execution role here, the big-picture view and the ability to evolve the system matter as much as the code itself.

1. 30 min call with Talent Acquisition

2. Technical assessment 

3. Vision & systems interview with the Chief Data and AI Officer

4. Team meet-and-greet