Philosophy, linguistics, and German Studies at HHU Düsseldorf — Wittgenstein, pragmatism, philosophy of language. I went in expecting academic philosophy and came out thinking about how meaning actually works in practice: how framing shapes what's possible, how language constrains and enables thought, how the same problem looks completely different depending on how you describe it. That background didn't lead to a philosophy career — it became the operating system. I default to asking what follows from different framings rather than chasing fixed definitions.

I work full-time at a digital marketing agency in Düsseldorf — 5.5+ years and counting. Started as a working student — Webflow builds, web design, concept development, copywriting, SEO. From there I moved into client consulting, strategy, and project management. The full stack of agency work. I taught myself to code along the way — HTML, CSS, Python, some TypeScript. About three years ago, AI changed how I work. Not as a shortcut — as a multiplier. The strategic thinking, the client relationships, the creative judgment — that's still mine. AI handles the execution bottlenecks that used to limit what one person could deliver.

Now I'm building agents — with the Anthropic Agent SDK for cloud, and with local LLMs through the OpenAI Agents SDK and Ollama. The focus is on agentic systems that deliver real value: not demos, not proofs of concept, but workflows that scale what one person or one team can actually do. I think the prompt is becoming the fundamental unit of knowledge work — and the interesting question is what happens when you build systems around that idea.

I'm sharing what I learn along the way — the strategy, the builds, the things that break — and documenting what practical AI work actually looks like.

The direction: self-hosted AI infrastructure, privacy-focused setups, agents-as-a-service that get things done on your behalf. I'm setting up a Linux home lab, learning in public, and working toward a future where useful AI isn't locked behind platforms you can't control. Agents talking to agents, running on your own hardware, serving your actual needs.