Roadmap AI per le Aziende del Settore Education & Training a Maribor

To modernize the University of Maribor’s legacy educational structures, we propose a three-tier AI integration framework. First, 'Administrative Offloading' uses LLM-based agents to handle multilingual student inquiries (Slovene, English, and German), which are frequent given Maribor's proximity to the Austrian border. Second, 'Curriculum Synthesis' employs AI to scan regional labor market data from the Drava Statistical Region, automatically suggesting course updates to match local industrial demands in automotive and chemical engineering. Finally, the implementation of RAG (Retrieval-Augmented Generation) systems localized to Slovene ensures that academic research remains accessible and searchable across all faculty databases without compromising data sovereignty.

A significant challenge for AI transformation in Maribor is the performance of global LLMs on the Slovene language, which often lacks the nuance of higher-resource languages like English. Our strategy involves fine-tuning open-source models (such as Mistral or Llama variants) specifically on Slovenian legal, educational, and cultural corpora. This prevents 'hallucinations' in legal training modules and ensures that AI-generated educational content adheres to the Slovenian Ministry of Education's specific pedagogical standards. Furthermore, we address the 'Digital Divide' by ensuring low-latency deployment of these models, making them accessible to smaller private training centers across the Podravska region that may lack high-tier compute infrastructure.