"Deep down, I'm still a scientist. I like understanding things properly. I like systems that behave predictably, and complexity that earns its keep."
Ask why patterns exist before deciding to change them
Only add complexity that justifies its existence
Classical ML or LLMs. Whatever the problem actually needs
From microscopy to SAP landscapes
During my PhD, I worked on machine learning for super-resolution microscopy. Early on, my instinct was to build things from scratch—to understand every moving part by re-deriving it myself.
Over time, the more interesting work happened when I stopped asking "how do I replace this?" and started asking "why does this pattern exist in the first place?" Combining, adapting, and slightly improving existing ideas turned out not to be a lack of originality, but how complex systems actually grow.
I learned to pay attention to which ideas survived repeated use, where they broke down, and how small changes propagated through the system. In practice, novelty often came from understanding rather than invention.
"Understanding accumulated solutions is a form of originality."
Today, I apply the same way of thinking when building AI systems inside existing enterprise landscapes:
Running on BW/4HANA
Via Cloud Foundry & SAP AI Core
Tools, feedback loops, deployment paths
LoRA, chosen pragmatically
Deep down, I'm still a scientist. I like understanding things properly. I like systems that behave predictably, and complexity that earns its keep.
For now, I build things that run, learn from the parts that don't, and keep choosing problems that are difficult for the right reasons.
How I balance work, training, and recovery
What I did and when.
Coding agents that write and test ABAP on BW/4HANA. RAG over SAP data on SAP AI Core. LangGraph agents with tools, feedback loops and deployment paths.
Training analytics for endurance athletes: performance prediction, Strava and Garmin integration. Built and run by me alone.
University of Würzburg - Developed ReCSAI and other AI algorithms for super-resolution microscopy. Published 15 papers with 442 citations.
Biophysics, image processing and early machine learning work. First papers on correlative microscopy and image registration.
Core competencies across the AI development stack
What triathlon has to do with my work
Triathlon is where I practice patience. Periodization, high-volume weeks, recovery weeks, and the discipline to stick to the plan when fitness isn't measurably improving.
Racing Ironman distance taught me to keep working through long stretches without feedback. That is most of what shipping a system actually is.
European Championship
German Triathlon League
1203km, 17700m elevation
Sports analytics platform
Research presentations and speaking engagements
Conference Presentation
Team Collaboration
Award Ceremony
If you have a problem that is difficult for the right reasons, I want to hear about it.