AI Automation Engineer (f/m/d)

Mission

Integral was founded in October 2024 and is already one of the fastest-growing startups in Germany. We're building the leading AI-native platform for accounting, payroll, and tax for European SMEs, and we're looking for the sharpest people to help us get there. The growth engine is running. Now we need someone to make it faster, smarter, and impossible to slow down.

We're hiring one engineer whose job is to use AI to automate manual work across every team at Integral and make every employee dramatically more productive. You will work directly with the Founders.

The way to build and run software products has changed. We call our approach SDLC2 — Research, Plan, Implement, with agents as first-class teammates and "agents as code" managed in one repo with PRs. We've written about it publicly at engineering.integral.de.

That same shift is now coming for the rest of the company.

Your Tasks

• Ship one working automation per week. Pick the highest-friction workflow you can find. Build the agent. Measure the hours saved. Move on.

• Define/enrich/upate sources of truth across Integral. Make our knowledge — SOPs, KBs, code, client data — navigable by the agents that need it.

• Own Context Management. Deploy agents that detect drift, flag contradictions, and identify gaps between what's documented and what's real. Define the human processes that keep documentation and reality in sync.

• Own Integrations. Build and maintain the MCP servers and internal tooling that let agents access and act on our systems

• Own Agent Development. Build agents that do real work. Write real eval suites — 20–50 cases drawn from real failures, not toy examples. Monitor trajectories. Iterate.

Your profile

• 4+ years shipping production software. Full-stack TypeScript or Python. You're fast prototype to prod in days, not sprints.

• 1+ year building/thinking/shipping LLM applications. With evals, not just prompts. You treat eval.ts as a first-class file.

• You've built an MCP server — or something equivalent. You can walk us through it line by line.

• You live on the AI frontier. You try every model, every framework, this month. You have strong opinions on Cursor vs Claude Code, on RAG vs context-stuffing, on when to use an agent versus a script.

• High