Engineering
AI Engineer Agentic Workflows (m/f/d)
You build the AI specialist roles that run in live operation at our customers, for practice areas that rely on the result. Each role consists of individual processing nodes that can be swapped, extended and checked one by one, and every work step is logged. Autonomy here has guardrails.
What the role involves
- Build AI specialist roles from individual processing nodes, each node individually testable and replaceable.
- Set the guardrails within which a role decides for itself which additional source to draw on and when a case is put before a person.
- Maintain traceability down to the source: every figure in the chat, in the dashboard and in the report carries the document, the system and the date of collection.
- Connect sources, from competitor sites and PDF data sheets through patent and approval databases to the customer's CRM and ERP.
- Build the dashboard and chat of a role so that a responsible specialist can expand every figure down to the individual record.
- Swap models and measure the results against the previous version before a change reaches the customer. Testing is done with dedicated test cases per role, that is, inputs with an expected result that run through on every change; there is no bought evaluation tool, because the subject matter looks different at every customer.
- The stack: Python for the agents, Node.js for the services around them, both in containers. Orchestration is done with LangGraph, where each work step is a node with its own state, individually checkable and replaceable. As a model we use Claude from Anthropic, and we make the choice per task. State and results are held in PostgreSQL via Supabase, research runs through Tavily, and operation takes place in the European Union on servers in Germany. Where a customer requires it, processing runs entirely within their environment.
What you bring
- Several years of experience in software development, part of it with LLM applications that are in live operation and did not stay at the prototype stage.
- Confidence in Python, in clean interface work and in the orchestration of multi-step processes.
- The habit of measuring a result rather than judging it: test cases, regressions, comparison against the previous version.
- An interest in the customer's subject matter. Anyone who does not want to know how a price list is read in sales will build the wrong role.
- Confident German and English, because customer conversations are in German and part of the documentation is in English.
What we offer
- Roles that run in live operation, not demos: what you build is used by a practice area the following week.
- Early responsibility for your own client projects and direct feedback from the founders.
- A team that works with specialized AI roles: whatever can be delegated is delegated. Your time goes into clients, quality and direction.
- Hybrid work with an office in Hamburg's HafenCity, equipment of your choice, development through responsibility rather than a course catalogue.
- Partners such as Morningstar, EY and the Projective Group, with whom you work on projects.
Apply as AI Engineer Agentic Workflows.
Your CV and three sentences on why this role. We do not need a cover letter. We reply within five working days.
By e-mail to bewerbung@scoreprise.ai
Please send your CV and references as PDF. A cover letter is not required; tell us in the e-mail what interests you about the role.
We treat all applications regardless of gender, age, origin, religion, disability and sexual identity. How we handle your application documents is set out in the privacy notice for applicants German.