Senior AI Engineer, Agents and Platform (f/m/d)

Your tasks

Your mission

The role In this role you own the platform every anzel agent runs on, and you set its architecture: what we build, what we buy, what we retire, written down so people and agents work from the same decisions. At anzel, AI writes most of our code and engineers guide it, so you build alongside a small senior team, close to the customer.

The hard part isn't the demo. It's making agents dependable for many customers at once: correct answers, each customer's data kept separate, and work you'd trust with a real sponsorship deal. What you’ll own

• The harness. How a skill written in plain language becomes a reliable agent: tools, orchestration, long running work that resumes after a failure, memory and context, retrieval, structured output, model routing, fallbacks and rollback. Models and providers sit behind our own interfaces, so we never depend on one vendor.
• Tools and MCP. The tool layer our agents use, which tools each agent gets, how they're versioned, and how access is scoped and logged per customer.
• Safe agents and separate data, with our head of engineering, who owns infrastructure, security and compliance. Each customer's data is isolated, every tool has only the access it needs, prompt injection is defended against, and every agent's actions are recorded.
• Quality in the pipeline, evaluations run on every change. The test sets come from our practitioners and our evaluation engineer, so nobody marks their own work. Every run is traced, with the cost and time of each piece of work visible.
• The engineering system, continuous delivery, review with coding agents doing the first pass, and releases that roll back in minutes. You lead by building, with clear decisions and fast, kind reviews.

Your profile

• Eight or more years building and running production software, or the equivalent built on your own. Real backend depth, APIs, a relational database, queues, long running workflows.
• Language model products that real users relied on, you can tell us about one that failed and what you changed.
• Retrieval that works, agents beyond a demo, and the judgment to know when an agent is the wrong tool.
• Evaluations that proved whether a change helped, y ou distrust a passing test until you know what would make it fail.
• Coding agents every day, several at once, without shipping code nobody understands.
• A security mindsetfrom running systems that held customer data.
• Something you've built that you can show us.
• Professional English; German is a plus.

Why us?

• You own the platform of a new product, with real customers from day one.
• A small senior team, almost no bureaucracy, your own machine and your own choice of tools.
• The work matters, it helps sport close better deals, price with confidence and put more money back into the game.
• Startup energy with backing, You get the pace of a startup, supported by XU Group's experience and network.
• A strong package, highly competitive pay that reflects the seniority and responsibility of the role.

We show an excerpt and a link, and the full text when the source allows it. Brands and logos shown belong to their owners.

Source: Statistisches Bundesamt (Destatis), table; dl-de/by-2-0; StarIn's calculation