Mission
Preconstruction is the foundation of every project, where the work gets chosen, priced, and put together. But while the projects have grown, the documents multiplied, and the requirements gotten heavier, the way the work gets done looks almost the same as it did decades ago.
We are building Arctis AI to transform how preconstruction works, making the people who run it faster, sharper, and able to take on more than ever before. By partnering with leading construction companies, we are pushing the boundaries of applied AI and unlocking a multiplier for precon teams.
You'll ship real features alongside the founding team.
Document Processing: Help build and improve the pipelines that turn messy construction PDFs into structured, queryable data.
AI Agents: Contribute to our multi-agent system, build evaluation datasets, test retrieval strategies, improve prompt performance.
Evaluation: Help define and run evals that measure whether the system actually works (accuracy, recall, failure cases).
Product: Touch the full stack. Fix bugs in the frontend, add API endpoints, improve the experience for real users.
You're early in your career but you've already built things on your own such as side projects, hackathon wins, open-source contributions, or just apps you made because you were curious.
You pick things up fast, ask good questions, and ship. Based in Munich or able to relocate for the duration.
Strong fundamentals in Python or TypeScript, ideally both
You've built something end-to-end before, even if it was small
Genuine interest in LLMs, AI agents, or NLP (bonus if you've built with them)
Comfortable with git, APIs, and figuring things out from docs
Familiarity with React/Next.js or FastAPI is a plus
Curious. You dig into things. You want to understand how it works, not just that it works.
Self-driven. You don't wait for instructions. You see something that could be better and you try to make it better.
Low ego. You're here to learn and contribute. You take feedback well and give it respectfully.
Real work, real impact. You'll contribute to the product our customers use.
Learn from builders. Work directly with the CTO and founding engineers. See how an early-stage startup actually operates.
Competitive internship compensation.
Whatever tools you need. Hardware, Claude Code, unlimited AI credits, productivity setup — just ask.
EGYM Wellpass, team dinners, and occasional offsites.
Intro call — Meet a founder. (~30 min)
Technical deep dive — We go deep on your experience, how you think about systems, and how you'd approach real problems in our domain. (~1 hr)
On-site day — Spend a day at our Munich office at Sendlinger Tor: work through a practical case based on our actual work, and build alongside the rest of the team.
Reference checks — We talk to a few people you've worked with.
Offer.
Search for a command to run...