About me
I’m a software engineer, researcher, and product builder based in Vienna.
I’ve spent over three decades writing software across industries — telecommunications during the GSM boom in Austria, surgical navigation systems in medical tech, eight years at NIM (Nuremberg Institute for Market Decisions), Chief Founding Engineer at ZML building a Zig-based AI inference framework focused on performance, and now running my own company: AI Research & Technology Lab GmbH.
What I’m Building
QualCode — AI-powered qualitative coding for researchers. It runs two LLMs as independent raters (OpenAI + Anthropic), automatically computes inter-rater reliability (Cohen’s kappa, Krippendorff’s alpha), and produces audit-ready output. Built for the dual-coding standard that serious qualitative research requires. GDPR-compliant, EU infrastructure.
LaBrowser — A research browser for behavioral studies. Captures structured interaction data — navigation, queries, AI prompts and responses, clicks — from study participants, with a researcher console for study design and analysis. Addresses a gap that screen recording and think-alouds can’t fill.
Background
I’ve been building things at the edges of disciplines for a long time.
At NIM I spent eight years as a senior researcher and technology lab manager — doing research myself, and building the infrastructure that made otherwise impossible research possible. That included a fully automated GPT-4 market research agent (SurveySeed), an Arduino-based sensor synchronization device for multi-seat biosignals labs, reaction time measurement tools for behavioral economics, online shop environments for controlled studies (web and native mobile), and a research browser prototype that later inspired LaBrowser — among other things. I also did a significant amount of AI research: image models, voice models, multimodal models, training, fine-tuning — a broad range of applied work.
In open source: I created ⚡ZAP⚡, the first production-ready HTTP web framework for Zig, which I presented at SYCL 2023. Telekasten.nvim started as my own note-taking setup in Neovim and ended up with a few thousand users. Before that, Sublime_ZK did the same for SublimeText — and when people asked for a version that didn’t require SublimeText, I built Sublimeless_ZK, a standalone Zettelkasten app.
I’ve also built a CPU from scratch, trained convolutional neural networks for logo detection in 2017, and written not one but two presentation systems — first a Godot game engine-based slideshow, then Slides in Zig (presented at SYCL), which has since been ported to Raylib as RaySlides.
The common thread isn’t the domain. It’s seeing a gap — usually from being close enough to a real problem to feel it — and then building from first principles rather than just reaching for the obvious tool. I didn’t build a Kanban board. I built OMS. I didn’t build another slide deck app. I built a presentation system on a game engine, then rewrote it in Zig, then ported it to Raylib. That’s just how it goes.
How I Work
Almost entirely in the terminal. Neovim, tmux, command line. Zig and Python are my primary languages, though I’ve written production code in C, C++ and enough others to get into trouble.
Since late 2025 I’ve been doing agentic engineering — Claude Code fundamentally changed how I work, and the shift is real enough that I’ve been thinking and writing about it. Running a one-man company forces you to think carefully about leverage: what compounds, what doesn’t, and how to build systems that work for you rather than the other way around. That thinking shows up in the blog too.