Interloom is a Berlin-based startup. We are building the first navigation system for work (think about of it as Claude Cowork for Enterprises). We are now looking for an AI Engineer to join our AI Engineering team. We are flexible on level. This role is a strong fit if you have solid SWE fundamentals and a deep interest in building production agent systems. Work from the Berlin office, hybrid or remote (+-1 hour from CET).
Most companies run on operational knowledge that never fully makes it into documentation: decisions hidden in tickets, emails, workflows, internal tools, and the heads of experienced employees. Interloom captures how expert teams actually resolve work and turns it into a living context graph - a memory layer that lets AI agents act on a company's real, proven decisions instead of generic training data.
We are already working with enterprise customers such as Zurich Insurance, JLL, and Fiege across facility management, insurance, banking, logistics, energy, and ITSM. In March 2026, we raised a $16.5M seed round led by DN Capital, with participation from Bek Ventures and Air Street Capital.
About the role
You will work end-to-end on the systems that make our AI agents capable, reliable, and safe in production. This is practical engineering work close to real users and real enterprise workflows. The details matter: context building, tool design, retry and recovery paths, observability, and evaluation quality often determine whether an agent actually works in production.
What you will do
- Build and refine agent capabilities through tools, skills, and integrations that are safe, observable, and easy for LLMs to use correctly.
- Improve the agent runtime: tool calling, orchestration, error handling, handoffs, and recovery paths.
- Identify agent failure modes and improve reliability so agents fail safely instead of silently.
- Build and improve evaluation frameworks: scenarios, graders, and tooling for testing agents across multi-step workflows.
- Work closely with product and customer problems, not only isolated research tasks.
- Help shape how production-grade agent systems should be built.
What we are looking for
- Strong backend engineering experience, ideally with Python.
- You write tested, type-checked, maintainable code.
- Solid understanding of relational databases, especially Postgres/SQL.
- Good API design skills.
- Interest in LLMs beyond prompting: tool calling, context design, evaluations, observability, and failure modes.
- Ability to take a bounded problem, make a small plan, ship the change, and communicate tradeoffs clearly.
- Comfort working close to product problems and real users.
Nice to have
- Practical AI/LLM engineering experience.
- Experience with context engineering, RAG, agent frameworks, evaluations, or LLM observability.
- Experience with Temporal or other durable workflow engines.
- Experience using metrics and evaluation data to improve system quality.
- Experience building reliable systems for enterprise customers.
How we work
- Small teams with real ownership.
- Minimal bureaucracy.
- We ship working software to users early.
- We value precise thinking, plain language, and written context.
- Humans stay in control; agents handle routine work so people can focus on decisions that matter.
Compensation and benefits
- VSOP, so you share in the upside you help build.
- Pick your own workstation and hardware.
- Comfortable offices in Munich and Berlin.
- Hybrid work if you are in Berlin; remote is possible otherwise.
- Annual company retreats.
- High-trust engineering culture.
- Clear growth paths as the company scales.
Location
Berlin hybrid or remote. If you are in Berlin, you can work from our office at Leuschnerdamm 13. Otherwise, remote work is welcome as long as your working hours align with Central European Time. Occasional travel to the team and offsites may be needed.