Senior AI Engineer, Agents and Platform (f/m/d)
Own and architect the platform every anzel agent runs on — an AI product that produces commercial work for the sports business (sponsor ROI reports, pricing analyses). Day to day you build the agent harness (orchestration, retrieval, memory, model routing), the tool/MCP layer, data isolation and eval pipelines, plus the CI/CD engineering system, alongside a small senior team.
- Compensation
- Not disclosed
- Location
- DE
- Work setup
- Remote
- Seniority
- Senior
- Employment
- Not specified
- Experience
- Not specified
- Category
- Ai engineering
- Posted
Skills
Must have
- 8+ years building and running production software, or equivalent
- Real backend depth: APIs, relational database, queues, long-running workflows
- Language model products that real users relied on, including learning from a failure
- Retrieval that works and agents beyond a demo, with judgment on when an agent is the wrong tool
- Evaluations that proved whether a change helped
- Uses coding agents every day, several at once, without shipping code nobody understands
- Security mindset from running systems that held customer data
- Something you've built that you can show
- Professional English
Nice to have
- German
Full description
Your mission
Join anzel - The AI for the business of sport.
Clubs, leagues and federations hold deep commercial knowledge, but most of it sits in spreadsheets and in the heads of a few experts. anzel turns that knowledge into finished work: sponsor ROI reports, pitch narratives, pricing analyses, partner fit assessments.
Sports business experts who have done this work in sport write the skills in plain language. Agents do the work over a knowledge base those experts have checked. Nothing reaches a customer until it passes its tests. Think of anzel as an experienced senior colleague who has done this many times before.
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. APIs, a relational database, queues, long running workflows.