AI Engineer

pAgents, knowledge graphs and evals: on data that disagrees with itself. /p ul liWork model: Hybrid, 2–3 office days a week /li liType: Full-time /li /ul pIn 1999, NASA lost a $125 million Mars orbiter because two teams meant different things by the same number. Lockheed Martin's software sent thruster data in imperial units. NASA's navigation software read it as metric. Both ran without errors. They just didn't agree. /p pMars Climate Orbiter
· 1999 $125M orbiter lost /p pEvery company has its own version of this. A "customer" in the CRM isn't quite a "customer" in the ERP, and most of the rules were never written down. Now that AI acts on that data, those mismatches get a lot more expensive. /p pG*rtn*r named context layers a $69 billion opportunity in August 2026. /p pAt Metagem, we're building the context layer for the AI era: agents that read an enterprise's documents and data, extract its definitions and rules, and surface where things conflict and what to fix first. That context powers every BI tool, assistant and agent the company runs. /p pWe're looking for an AI Engineer to help define this category with us. /p h3Why it's hard /h3 pThe data is messy, and language models won't always give the same answer twice. Your job is to turn both into something a business can rely on, from the first experiment to production. /p h3What you'll do /h3 h3Build agents /h3 h3Measure what works /h3 pDefine what good looks like, measure it, and tell a real improvement from noise. /p h3Keep the knowledge graph honest /h3 pModel ontologies, merge duplicate concepts, and keep the graph consistent as documents and data change. /p pBuild assistants and tools that answer from governed context, through channels like MCP. /p h3Ship it /h3 pBuild production services and deploy them into customers' own cloud environments. /p h3Work with customers /h3 pSit with their data teams to understand the problem and check what the system finds. /p h3Who we are /h3 pMetagem is building the agentic context quality layer for the European enterprise. We're an early-stage company, founded in 2025, with roots in academia and data consulting. Our customers are industrial groups with a lot of data complexity. /p pBeing this early means you'll work directly with the founders across AI engineering, software architecture and business strategy, and you'll shape product decisions from your first week. It also means not everything is figured out yet, so you'll help decide how things get built. /p pAnd you'll grow with it. As we scale across Europe, early engineers take the lead: owning major parts of the platform, setting the technical direction, and building and leading the engineering team that comes after them. /p h3Who you are /h3 pYou want to build something that matters, and you want to build it now. We're looking for an exceptional engineer who sets a high bar, moves fast, and takes full ownership, from the first prototype to what runs in production at some of Europe's largest industrial groups. /p pIn practice, that means: /p ul liYou have a degree in Computer Science or Engineering, or you've proven the same by shipping real things. /li liYou write excellent Python and you've taken code all the way to production. /li liYou've built with LLMs (agents, RAG, embeddings or structured extraction), and you insist on measuring whether they actually work. /li liYou're strong with data: SQL, data modelling, and making sense of a schema you've never seen within an afternoon. /li liYou hold yourself to high engineering standards: Git, testing, clean code, Docker and cloud deployment. /li liYou use coding agents to move faster, and you review their output as carefully as a colleague's. /li liYou think in systems and you want the hardest, most open-ended problem in the room. /li liYou explain complex ideas clearly in English, including to executives who aren't technical. /li liKnowledge graphs, ontologies or graph databases /li liAzure, AWS or Google Cloud data and AI platforms /li liAgent frameworks and MCP /li liFine-tuning or running open-weight models /li /ul pIf you want your work running inside Europe's largest enterprises, not sitting in a slide deck, you'll fit right in. /p h3Our stack /h3 ul liPython /li liFastAPI /li liAgent frameworks /li liNeo4j /li liQdrant /li liMCP /li liDocker /li liReact + TypeScript /li /ul h3What you get /h3 ul liA role with real impact: you own a core part of the product end to end /li liA training budget (certifications included) and access to the latest models and tools /li liCompetitive pay and early-stage equity (ESOP), based on experience and performance /li /ul pbStart date: /b flexible. bLocation: /b Wintercircus, Ghent. Hybrid, with 2 to 3 days a week in the office. /p h3How we hire /h3 ul li01pA 30-minute intro call to get to know each other. /p /li li02pA technical conversation about something you built. Bring something you're proud of. /p /li li03pA short exercise based on a real problem we work on. Wh