Software Architect · AI, Web & Cloud · Paderborn, Germany
I don't think our job is writing code anymore. Our job is building the factory that writes it.
At dSPACE I work with our development teams on an AI software factory: the setup in which AI agents turn requirements into working, tested software for a large enterprise, fast and at a cost that still makes sense at scale. My part is the architecture of that factory: the platform, the guardrails, the context agents get through MCP, and the process around it, from the first requirement to the reviewed pull request.
I also lead the company-wide Web & Cloud architecture community and write the guidelines our teams build on. Before AI filled my calendar, I spent more than 20 years with .NET, COM, containers and the cloud. So I care less about the newest model and more about what still runs in production a year later.
Some days the job title should read AI Factory Facility Manager.
- What does a software factory need so that agents deliver enterprise-grade results, not demos?
- Where does the cost go: tokens, review time, or rework?
- How do you review code when agents write most of it?
- What belongs in a requirement before any agent starts working?
- Does spec-driven development hold up in a team of hundreds?
- AI software factory: agent harness, skills, MCP, RAG, guardrails, cost control
- AI software architecture: AI-ready products, agentic workflows, spec-driven development
- Web & cloud: Kubernetes, Azure, .NET, TypeScript, React
dscom: exports .NET assemblies to COM type libraries,
a replacement for tlbexp.exe and TypeLibConverter. Maintainer.
iSAQB CPSA-F and SWARC4AI (architecture for AI systems) · CKAD · founder of the .NET User Group Paderborn · speaker at dotnet Cologne
AI and agentic engineering
Cloud and platform
Languages and web
