SecureRAG
With LLMs and RAG, we enabled our industry partner to let authorized users query internal databases in plain language—simplifying access to key project and employee data.
Factsheet
- Schools involved School of Engineering and Computer Science
- Institute(s) Institute for Cybersecurity & Engineering (ICE)
- Funding organisation Others
- Duration (planned) 01.03.2025 - 31.08.2025
- Head of project Prof. Dr. Ulrich Fiedler
- Partner Edorex AG
- Keywords Artificial Intelligence (AI), Large Language Models (LLM), Retrieval-Augmented Generation (RAG), Natural Language Querying, SQL Generation, Knowledge Access, Data Security, Proof of Concept, AI-Supported Workflows
Situation
The industry partner, Edorex Informatik AG, Ostermundigen, selected us as collaboration partners to build expertise in AI workflows. They aimed to develop a proof-of-concept for a secure chat system enabling all authorized employees to access internal project and employee data in plain language.
Course of action
The project followed an iterative approach: analyzing user needs and database structures, integrating LLMs with RAG, and building a secure, user-friendly chat interface tested and refined with real company data.
Result
The proof-of-concept showed that authorized employees can securely query internal data in plain language. The system generated valid SQL, delivered clear answers, and laid the foundation for future AI-supported workflows.
Looking ahead
The partner aims to strengthen their AI expertise, build a full-scale system, and gather hands-on experience with pilot users to guide future adoption.
