
Mycellium
Challenges the orchestrator model with fully decentralised agent collaboration
Role: DevOps & infrastructure owner · agent communication layer
- Python
- Docker
- Railway
- CI/CD
I work where applied AI meets dependable infrastructure. I design the system, own the delivery, and stand behind it after the demo ends.
Three systems I carried from a blank repository to something other people now depend on.

Challenges the orchestrator model with fully decentralised agent collaboration
Role: DevOps & infrastructure owner · agent communication layer

Centralized resource hub for students
Role: Information architecture · full front-end · data layer

Live production store serving real customers
Role: Full-stack build, end-to-end · AI assistant integration
Three things I keep coming back to, learned the slow way.
A system someone else has to decode isn't finished. The best code reads like the answer was obvious all along.
Anything can look good once. Trust is earned in the maintenance, the tests, and the years that follow launch.
Technology grounded in a genuine problem outlives technology built to impress. I'd rather solve the first kind.

I moved to Munich in 2021 to study Information Systems at TUM. Since then I've shipped code across ad-tech, retail IT, enterprise research with SAP, and a real release train at BMW. Each stop taught me a little more about what “done” actually means.
Read the full story