I build privacy-first apps — tools that run on your own device, work without accounts or a backend, and only reach the cloud when you bring your own AI key. No telemetry, no data collection, no lock-in.

Most of my time right now goes into TobiBoard, an on-device Android keyboard with optional bring-your-own-key AI voice-to-text and text fixing, and TobiReader, a local-first RSS reader with on-device read-aloud and optional AI summaries. Around those I keep a small portfolio of shipped apps — a cross-platform screen recorder, macOS voice-to-text and menu-bar utilities, and a few AI-assisted side projects. The full list is on the projects page.

The thread through all of it: see how far useful, AI-flavoured software can go while keeping user data on the user’s own hardware. In practice that’s two patterns — run models locally where it’s feasible (on-device speech and text-to-speech, local LLMs via Ollama on the desktop), and bring-your-own-key for the rest, so anything cloud-bound uses your API key and endpoint, never mine.

I also do some consulting through Leinss Consulting, helping service companies put generative AI to work — but building my own apps is the main focus these days.

Based in Germany. This is a now page; last updated June 2026.