AI Engineer

Most AI projects are demos. This one is live.

Munich · hybrid InternshipWorking StudentStudy Project Apply now

1 to 4 CS students
3 to 6 months
about 30 h / week
min. 50% on-site, Munich
Start ASAP
TUM IDP

Germany’s trades run on paperwork, and we are fixing that with AI. You join a live product with real pilot customers and build on a working multi-agent setup, never starting from zero.

What you’ll do

Extend the KAI orchestrator and its sub-agents that run a company’s back office.
Strengthen the knowledge graph and structured outputs so KAI answers reliably, with no hallucinations.
Help build KAI Voice, the phone agent that captures requests autonomously.
Fine-tune domain embeddings on real pilot data and sharpen the payment and DATEV integration.

What you bring

You code confidently in Python and have worked with LLMs, RAG or agents.
You think in systems and want things running in production, not just in a notebook.
A CS Master’s (ideally TUM CIT) fits the IDP, but your drive matters more than your degree.
AI-assisted development with Claude and Codex is a must, it is how we work every day.
English fluent, German a plus.

Stack

PythonLangChainpgvectorFastAPI

Why KalkulAI

Radical ownership from day one. No fixed job description, you own outcomes.
A live product with real pilot customers, not a lab hypothesis.
AI-native: Claude and Codex are part of how we work every day.
You sit side by side with the founding team.
Backed by EXIST, TU Munich and UnternehmerTUM.

How to apply

1 Apply with your CV, GitHub and a short note on which role and why, to [email protected]. Multiple applications are welcome.
2 A first call with the team, 20 to 30 minutes. Depending on the role, a small case or take-home task can follow.
3 After a yes, we set up the format together: as an internship or working student role, or we handle the TUM IDP registration with a supervising chair.

Backed by

Apply for AI Engineer

The steps are listed above. Questions before you apply? [email protected]