Vertical software for German industries, built with current AI models hi@westwoodworks.de
DE EN
About

About Westwood Works

A handful of master’s students with a startup. We build software for problems we see in people we know.

What we do

We’re a startup of master’s students. We build software for problems we see in the people around us: teachers preparing lessons, people who let apartments or develop residential property, a winery unhappy with its ERP. That became Klassenwerk, Edifondo and Pinoterra. Ziffero is the bookkeeping underneath, KontoKit the open-source companion.

Free on top: the Edifondo project calculator at edifondo.de, and Watch, a platform that tracks legislation and regulation. No market study as the starting point, no client work — and we’re far from done.

How we work

Quality over quantity. Every product gets the time it needs. No half-finished releases to hit an arbitrary date. What’s in beta is called beta.

Traceable. For AI features, “sounds plausible” doesn’t count — “matches the source” does. Every summary links its document. Every curriculum-bound output traces back to the official Lehrplan.

Direct. Support comes from the people who built the product. No ticket system, no hold queue. Replies usually within 48 hours.

Open. We write on the blog about what works and what doesn’t — technology choices, dead ends, evaluation methods. No marketing speak.

Why with AI

A general model knows a little about everything and not enough about anything to make it checkable. A model working against a known, finite corpus — all 16 Lehrpläne, the SKR03/SKR04 account frameworks, German wine law — can be tested: there’s a correct answer to measure against. That’s where we use language models. Not as a label, but because in a narrow field they take the actual work off your hands.

Technology

We choose technology by what fits the problem, not by what’s fashionable. Currently in use:

  • Backend: .NET, Go, Python, Node.js
  • Frontend: HTML/CSS, vanilla JS, Hugo for static sites
  • Models: Anthropic Claude API, OpenAI API, Gemini API — plus evaluation pipelines against domain-specific datasets
  • Infrastructure: Linux, Docker, simple deployments

Independence and longevity

Self-funded means no capital that can run out, and no investors for whom an exit matters more than the product. Anyone putting their bookkeeping or winery management on a piece of software wants to know it will still be here in five years. Our answer is the structure: small, debt-free, no outside pressure.

Contact

Questions about a product? An idea for what software should solve? Just curious?

Email: hi@westwoodworks.de
GitHub: Coming soon

Reply time: usually within 48 hours.