Senior AI Engineer

Joining Collibra's Artificial Intelligence team Collibra is the context engine that powers trustworthy enterprise AI, giving LLMs and agents the business context they need to reason accurately over an organization's data, while our governance platform ensures that same context, and the AI acting on it, stays governed, secure, and compliant. It's a two-sided problem: making AI smarter with context, and making sure that context and AI stay under control. This is not an internal data science or research role. You'll be building AI features that ship directly into the Collibra product used by real companies, including some of the largest enterprises and government agencies in the world, across every industry and continent. Real production AI: streaming responses, agent loops with checkpointing and resumption, running at scale for customers who depend on it daily. We need someone who combines strong software engineering with data science judgment to build it and keep it running. You'll work closely with Data Scientists, other AI Engineers, and MLOps Engineers, plus teams across Collibra, taking AI products from prototype to production and beyond. This is a hybrid role based in our Brussels office. Our hybrid model means you’ll work from the office at least two days each week. This setup helps us stay connected, work more closely together, and keep making progress as a team. Senior AI Engineers at Collibra are responsible for Shipping AI features into a real, live product not a prototype, not an internal tool used by companies around the world. Building, instrumenting, and evaluating AI/ML models and treat a quiet, no-output failure as a bug, not a shrug. Helping shape how Collibra's AI platform evolves, and mentor other engineers along the way You have 5+ years of software engineering experience building and running production systems, with 3+ years applying data science/ML in production (LLM applications, RAG, NLP, recommendation engines). Hands-on with agent-loop frameworks (e.g., LangGraph) and tool-calling standards (e.g., MCP): tool cycles, state across steps, human-in-the-loop interrupts, checkpointing and resumption. Experience with agent harness architecture, such as sandboxed code execution, tool permissioning, safely orchestrating autonomous actions. Experience with agent evaluation frameworks which test tool-calling behavior and agent trajectories. Strong observability instincts: a silent, no-output turn reads as a defect, not a non-event. Retrieval/RAG judgment: hybrid semantic + keyword search, rank fusion, precision/recall tradeoffs. Experience building stateless services at real scale multi-replica deployments, no sticky sessions, concurrent requests. Comfortable with SQL and relational databases, plus familiarity with time-series, graph, or vector databases. Strong command of modern Python typing/validation (Pydantic v2) and production AI tooling (LangChain, LangGraph, etc.). A bachelor's degree or equivalent related working experience is required. You are Deeply technical and analytical, driven by what the system is actually telling you, not the tidiest story. Able to cut through AI hype to find the real, shippable problem underneath. Excited by the architectural and mathematical challenges of building sustainable, governed value on LLMs. A clear communicator, equally comfortable in deep technical detail and the big picture, with technical and non-technical audiences alike. Product-minded: you think about how a real customer actually experiences an AI feature, not just whether it's technically impressive. A renaissance engineer at the forefront of agentic coding. You are fluent with AI-assisted development, but not a "vibe coder"; you treat these tools as leverage, not a substitute for understanding your own code, and you own the output and its cost. Curious and generous with what you learn nobody has this fully figured out yet, and we're building the playbook together. Measures of success Within your first month,…