As part of an extensive pilot project, a renowned Austrian university of applied sciences is evaluating the use of Mikis – My Individual AI Assistant for various application areas within the university context. The goal is to develop chatbots with specific functions for teaching, administration, and IT support, based on different language models – depending on data protection requirements and desired system performance.
Hybrid Model Approach: Cloud & On-Premises
The project combines both cloud-based language models such as ChatGPT and locally hosted models via the Ollama platform (e.g., Llama 3.2, Gemma 2). The local installation of Ollama was carried out on a GPU-supported server within a secure infrastructure to ensure high-performance response times even for complex queries.
This hybrid approach gives the university maximum flexibility:
While locally hosted models are used for internal and sensitive data, public or non-critical queries are efficiently handled via cloud-based services – always in a privacy-compliant and controlled manner.
Seamless Integration into the Existing System Landscape
Mikis was integrated smoothly into the existing IT environment. A key milestone was the connection to Microsoft Entra ID (formerly Azure Active Directory), enabling centralized user management and role-based access control. Access rights can thus be precisely managed for different user groups – from students and faculty to administrative staff.
Versatile Use Cases
Throughout the project, various chatbot scenarios were developed and tested, tailored to the university’s diverse operational needs.
Conclusion
This project clearly demonstrates how Mikis – as an open, modular AI platform – meets the complex requirements of a modern educational institution: technologically flexible, data-sovereign, and future-proof.
The combination of powerful language models, customizability, and high IT compatibility makes Mikis the ideal solution for universities seeking to implement AI in a responsible and strategic way.
