AI Engineer / Agentic AI Engineer - Remote (Germany) with occasional on-site days

We are a specialized AI engineering and full-stack team building productive LLM/agent systems for our customers — no PoC theater, no Powerpoint decks, no 'Slide Architects'. We advise partners collaboratively at eye level and deliver software that runs in production. You can solve problems here without being limited by titles or fixed role definitions — we don't have those. Our jobs are not rigid; they serve only as an orientation for you. You will not be confined to (one) job, but rather develop yourself across AI engineering, backend, platform, and consulting further. You stand with your individual goals, visions, and strengths at the forefront. You shape your career proactively at Arcode. Important: As we work closely with our customers in the DACH region (finance, insurance, public sector, industry), excellent German language skills (ideally mother tongue, at least C1) are absolutely required — conversational English is assumed. **What you can expect** As an AI Engineer (m/f/d), you will not only write notebooks and prompts; you will build productive GenAI systems: agents, RAG pipelines, LLM-driven workflows — end-to-end from use-case workshops to production deployment. You will work in cross-functional, agile teams on customer projects and come into contact with the entire modern GenAI stack: pydantic-ai, LangChain/LangGraph, MCP (Model Context Protocol), OpenAI, Anthropic Claude, Azure AI Foundry, AWS Bedrock. Together with your team, you will advise our customers on use-case identification, architecture decisions (RAG vs. fine-tuning, agent design, tool use, memory), effort estimates, and the selection of suitable process models — you will be involved in the entire AI lifecycle. You will analyze customer requirements, design and implement agent-based solutions, RAG pipelines over enterprise data, structured LLM outputs (pydantic schemas, tool use) and ensure good documentation. You will keep quality on your radar — and that means at AI: systematic evaluations, regression tests against gold sets, hallucination monitoring, cost-per-inference tracking, prompt versioning, observability for agent traces (OpenTelemetry, Langfuse, Phoenix). You will use modern, automated approaches and AI-native developer workflows (Claude Code, Cursor, GitHub Copilot) to optimize code, error recognition, and innovative solutions — we live 'AI-native', not just AI-for-customers. You will integrate your LLM/agent components cleanly into existing backend and frontend architectures (Python/FastAPI, Node.js/TypeScript, Java, React/Angular) and ensure clean handovers to the platform side (Kubernetes, ArgoCD, MongoDB/pgvector, AWS/Azure). **What you should bring** Thanks to your corresponding professional experience, you will pursue the AI engineering approach hands-on — you won't just talk about LLMs; you'll build with them. **Good to have** You are a team player, appreciate open communication, learn from others, and share your knowledge — especially in the LLM field, where the tooling landscape changes every three months. You want to work on exciting customer projects and have good knowledge of some of these technologies: AI/GenAI: Python, pydantic, pydantic-ai, LangChain/LangGraph, MCP, OpenAI/Azure OpenAI APIs, Anthropic Claude, Azure AI Foundry, AWS Bedrock, Hugging Face Transformers, PyTorch, Embeddings & Vector DBs (pgvector, MongoDB Atlas, Weaviate, Pinecone, Qdrant), RAG patterns, prompt engineering, agent-evals (Langfuse, Phoenix, ragas) Backend/FullStack: Python (FastAPI), Java/Spring or Quarkus, Node.js/TypeScript (Express, NestJS), React, Angular, MongoDB, PostgreSQL, REST/GraphQL Platform/DevOps: Docker, Kubernetes, ArgoCD, GitLab CI/GitHub Actions, AWS, Azure, MLOps, Observability (Prometheus, Grafana, OpenTelemetry) AI-native Tooling: Claude Code, Cursor, GitHub Copilot, Codium — you use them daily and can show customers how to use them productively. **Bonus, no must**: Own publications/Open-Source contributions/Konferenztalks in the AI field, experience with Machine Unlearning, Privacy Auditing, BaFin/EU-AI Act, or Edge AI. **Benefits** What we offer you: Whether remote or on-site — you decide how and where you work, as long as it's aligned with your project and team. Default is remote, on-site days at the customer's site (typically every two weeks). Off-Project Time — Most of your time will be spent working on customer projects. Approximately 20% of your working hours are available for further education, research, Open-Source contributions, conference talks, community engagement, or internal AI initiatives — we actively publish (ICPR 2026, IPDPS 2026, NeurIPS 2026 in preparation) and want you to be able to participate. You enjoy project and customer diversity across finance, insurance, public sector, and industry and work independently and self-organizing with the latest LLM and agent technologies in cross-functional teams. You get to choose your entire hardware including OS yourself — private use included. Plus a dedicated AI tooling budget (Claude/ChatGPT/Cursor/Copilot Pro licenses, API credits for experiments) — you should work with the best tools, not the cheapest. You find yourself in an environment that values family-friendliness, trust, freedom, knowledge sharing, participation rights, and a culture of error tolerance — in a field where 'making mistakes' is the only way to get GenAI systems running productively. In addition to flexible trust-based working hours, you have 30 days of vacation, special leave, accrued vacation time, and Christmas and New Year's Eve are generally not workdays. You can also take a sabbatical. For freelancers: 80-110 €/h depending on profile, clear project assignment, fair forecast communication.