About our projects
Studierendenwerk Ulm Chatbot
Commissioned by Studierendenwerk Ulm, we develop and operate the helpR chatbot.
The project focuses on the multilingual delivery of website information and the secure hosting of the Large Language Model (LLM) on a dedicated, in-house server infrastructure.
Technologies used
About the Studierendenwerk Ulm
Custom Knowledge Architecture
Rather than relying solely on generic vector search, we utilize a targeted and structured approach to organizing website content. By categorizing information into logical thematic groups, the chatbot can pinpoint relevant data with high precision. To ensure verifiability, every substantive response is accompanied by a direct link to its source on the website, significantly enhancing data traceability and transparency.
Data Sovereignty Through Self-Hosting
Data protection is a core requirement for public institutions. By self-hosting the language model on our own infrastructure in Germany, we ensure that no data leaves the European jurisdiction. Since no requests are transmitted to external API providers, GDPR compliance is intrinsic and an integral part of the architecture.
Cross-lingual communication
While the Studierendenwerk Ulm website is primarily available in German and English, the chatbot extends the linguistic range to include a wide variety of additional languages. By leveraging the German-language knowledge base through cross-lingual capabilities, the model can respond directly in the user’s preferred language. This allows international students to receive reliable answers in their native tongue without the need for the extensive manual maintenance of multilingual content.
CORTEX offers us tailor-made solutions, demonstrates the necessary flexibility to meet our diverse requirements and impresses with its reliable implementation. The company places great value on innovative advice, which is of crucial importance to us.
Claus Kaiser, Managing Director Studierendenwerk Ulm
