Want to build AI agents that actually give trustworthy, explainable answers not confident guesses? This is a role for a GenAI engineer who knows that the way you get there is by grounding LLMs in real, governed enterprise knowledge.
You’ll join a specialist team at a global life-sciences organisation building a new generation of knowledge-graph-powered AI agents. Your focus is the tooling and the applications on top of the intelligence layer, not the underlying data.
What You Will Build
Picture an agent that answers a question like “which priority hospitals in the US have decreasing sales?” To do it, the agent reads the definitions of the business terms from a knowledge graph so it understands the question, then queries the data warehouse for the real figures, and composes a grounded, traceable answer.
That grounding is the whole point: it’s what cuts hallucination and makes every answer explainable. If building that kind of system sounds like your idea of a good problem, read on.
What You Will Do
- Design and build LLM-powered agents and retrieval solutions on top of enterprise knowledge and data
- Connect agents to enterprise systems through tool definitions and MCP-style connections
- Benchmark and evaluate models, then take solutions from prototype into production
- Build reusable frameworks and accelerators for agentic AI
- Define the testing, evaluation, monitoring and governance for what you ship
What You Will Bring (essential)
- Strong hands-on GenAI / LLM engineering – you’ve built real solutions with LLMs: agents, RAG, prompt and tool design, benchmarking, and shipping to production
- Hands-on experience with knowledge graphs and semantic web in applications – SPARQL, RDF and related standards
- Strong Python and modern API development
- Solid software engineering fundamentals — Git, CI/CD, testing, cloud-native architecture
- A clear communicator who works well with both technical and business stakeholders
4+ years of AI engineering experience is a starting point — we care far more about genuine depth building LLM-powered systems than years on paper.
Nice To Have
- MCP (very learnable if you know LLMs and Python)
- Vector databases, embeddings and semantic search
- Any graph or semantic tooling — Neo4j, Stardog, metaphactory, Snowflake and similar (current set up is standards-based and vendor-neutral, so the standards matter more than any one product)
- Life sciences, pharma or other regulated-industry experience
The Details
- Contract role
- EU-remote based
- Occasional on-site workshops in Germany (roughly every couple of months)
- Start: 1st October
- Runs to year-end initially, with strong potential to extend into a full project in the new year
Apply Now
To explore this opportunity further and learn more, click the enquire today button.