Designed and developed an open-source AI system for intelligent search and interaction with institutional knowledge. The project covers the full system architecture and implementation, with a focus on reliable AI responses, accurate source attribution, data ownership, and independence from individual AI providers.
The system is used at the Museum of Prague and was featured at the Frankfurt Book Fair as part of the Moravian Library in Brno’s presentation.
Books, websites, curatorial texts, and other institutional materials can be made accessible through conversational AI. Rather than relying solely on a model’s general knowledge, responses are grounded in the institution’s own sources and accompanied by precise citations, making the results more accurate, transparent, and verifiable.
The architecture is modular and provider-independent, supporting OpenAI, Google, eInfra, or private infrastructure without locking institutions into a single commercial provider. Source materials can remain entirely on the institution’s own servers, preserving control and ownership of their data.
Built as an open-source solution, the system is intended for adoption and adaptation by other museums and cultural institutions for their own collections and workflows.
Role & contribution: concept, system architecture, AI/RAG pipeline, development, implementation, and open-source design.
