AI systems for organisations that work with their own documents. Built in Germany, run where the customer decides, with the data route written down before the first request goes out.
Dr.-Ing. Alexander Brunker · AI Solution Architect · Hückelhoven, Germany
- AMANA, a knowledge system for files that stay in the house. Retrieval over your own documents, a source named for every answer, pseudonymisation before every model call, and the choice of where the model runs. amana.leopardcode.ai
- Consulting and engineering for RAG pipelines, on-premises models and GDPR-native architectures. leopardcode.ai
- Training: a one day workshop on the EU AI Act, with your own documents and a file note at the end.
We write about what we build, with sources and dates, not with promises.
- RAG explained: a source instead of a guess
- AI models on-premises: what is GDPR-compliant about it, and what is not
- Pseudonymisation before the model call: why fail-closed counts
- Client data in ChatGPT and Claude: the risks, and the route with AMANA
- Using frontier models under the GDPR: four routes, one principle
- Camp brain: 150 people, no signal, one list
Everything in German too: blog.leopardcode.ai
