Software Engineering

Cologne/Bonn or remote

The invariants remain. The tools get better.

I build custom software for large organisations, for more than 30 years and most recently with AI agents. Agents shift the bottleneck: from writing code to specifying and verifying it. I have been doing this work since I wrote my first code generator.

I am looking for a long-term collaboration: a permanent position or a longer engagement as a freelancer, in Cologne/Bonn or mainly remote.

Approach in seven steps

Delimit the context
Fix the domain language
Formalise the model
Record decisions
Generate, by generator or agent
Verify and validate
Document and iterate

Each step has an artefact and a check. The details are under Approach.

What stays, what changes

Invariant	What it means when working with agents
Domain language	The agent's instruction file is a binding glossary.
Model as the source of truth	Generation works against a grammar, not against prose.
Context boundaries	The context window is a bounded context.
Decisions	Humans make architecture decisions, the agent implements them.
Verification	Every output has a check.
Iteration in the team	Review is a fixed step, not an exception.

Generator line

1997–2000 Select Software Tools (vendor of UML tools): automated data preparation from the UML tool SELECT Enterprise into Word, Lotus Notes, Cold Fusion; SQL generators
2004 Code generator for test drivers in a load-testing project
2008–2009 Model-driven migration of an ERP system (MDA)
2010–2014 CRM portal (telecommunications): unified SQL generation from Excel specifications, coding conventions, machine-checked quality rules
2014–2015 B2B interface (telecommunications): migration of a generator framework to Xtext, modelling workbench, generators for WSDL/XSD, verification by AST and text comparison
2016–2018 Information system of a federal authority: in-house tool for semi-automated UI migration, automated migration tests
2019 Preliminary study on data exchange between authorities: context analysis with a uniform DSL, XSD → Ecore, canonical event format
2021–present Systems for plausibility checking and planning at a federal authority: model-driven product line, specification language for rule-based planning, architecture documentation
since 2026 bewirb.es: three DSLs, code generation, M2M/M2T, built in dialogue with an agent
since 2026 Specification-driven development with AI agents (AI Unified Process), extending a legacy system of a federal authority with a new business area: vertical slice as a requirements elicitation tool, corrections only in specification and guidelines, never in generated code; rule set formalised from a sister service; agent environment opencode and Claude Code

All stages

Working together

Introducing agentic development in enterprise teams: control layer, conventions, verification.
Domain-specific languages and metamodels as a specification language for agents.
Modernising large Java legacy systems with agentic support.
Contributing as a senior engineer in the Java ecosystem.

Background

Diplom-Physiker (German degree in physics, University of Cologne; theoretical solid-state physics, minor in mathematics). Research on the theory of stochastic Hamiltonians at Forschungszentrum Jülich; published in Physical Review B 46, 5246 (1992). Computer science courses at FernUniversität Hagen.

Current interests: statistical physics of LLMs, quantum computer science, category theory.

Tailored profile

Send me your job description by email. I will generate a tailored profile as a PDF: from the same profile model as this website, with the project experience that matches your requirements.