About me
After 15+ years
in software engineering, I've turned my focus to the next frontier: intelligent systems.
I still advise startups on technical strategy, but these days I'm equally fascinated by the challenge of making AI agents actually work together. It turns out the hard part isn't building one smart agent; it's getting fifty of them to coordinate without chaos.
I'm the kind of engineer who cares as much about how the pieces fit together as what each piece does.
What I do now
Most of my time goes into the AI/ML space: building with PyTorch, LangChain, LangGraph, vLLM, and RAG systems. Prompt engineering, fine-tuning, agentic architectures. It's the most interesting technical frontier since the early web, and I'm here for it.
But I still get my hands dirty across the full stack:
- Frontend: React, Vue.js, TypeScript, Next.js
- Backend: Python (FastAPI, Flask), Node.js, Go, Rust
- Infrastructure: AWS, GCP, Azure, Kubernetes, Terraform, GitOps
- Data: PostgreSQL, Redis, Elasticsearch, vector databases
I care about developer experience, clean APIs, and systems that don't wake you up at 3am.
Track record
- Built the AI platform at Samsa.AI: multi-provider ML pipelines (fal, Replicate, HuggingFace), GPU infrastructure with PyTorch optimizations, and a cloud-native stack across AWS, Cloudflare, and Supabase
- Built auteon from a blank editor to production, passing due diligence and securing seven-figure seed funding in 6 months
- Took a PostgreSQL query from 15 seconds to under 500ms at PeterPark (that one felt really good)
- Led the frontend reboot at Sixt that replaced a PHP monolith with a React SPA serving 1M+ monthly bookings
I've been lucky to work across wildly different domains: automotive spare parts, parking IoT, EdTech, marketing automation, and AI/ML. The common thread? Complex systems that need someone to care deeply about how all the pieces fit together.
How I got here
Me in 1991 hacking on a Commodore PC-30. That was a 12 MHz 80286-based system including a 20MB hard disk, a monochrome text card, 512k of memory on the motherboard and two 5 1⁄4-inch floppy disk drives. I did play a fair amount of Tetris and Digger on this one.
Since i got my i486 DX with 33Mhz, a 105MB hard disk, a VGA graphics card with 512k and 4MB RAM in 1993 i had access to the internet. Besides fiddling with MS DOS and Windows 3.11 i played way too much Commander Keen. Although i started using a mouse back then to play Day of the Tentacle and The Secret of Monkey Island, it should take until 1995 that i upgraded it with a CD-ROM 2x Speed to play the latest games like Command & Conquer and Descent.
The next years i spent writing code, thinking about the best tools available to write code, reviewing other people's code, debugging, planning features, talking to others about all those things, and so forth.
Working with me
I tend to get INTJ-A or INTP-A in MBTI tests, which makes sense as i am a huge fan of Tim Urban.
My network describes me as an efficient, reliable, structured, creative, adaptive, conscious and inspirational engineer possessing technical depth and breadth.
When it comes to process, I default to lean and iterative: design thinking to find the right problem, agile to ship solutions, and enough structure to keep a team aligned without drowning in ceremonies.
Languages: German (native), English (fluent).
Let's connect
I'm open to AI engineering roles, CTO positions, and technical advisory work. Remote-friendly, interesting problems preferred.
"Avoiding complexity reduces bugs."
- Linus Torvalds