Implementation of an AI-Based Interactive Kiosk with Gemini API and RAG for Campus Information System Optimization

Abstract

Efficient and accurate information access is a fundamental challenge in higher education environments. This research aims to develop an interactive kiosk application based on artificial intelligence (AI) as a solution to optimize the information system at Universitas Kebangsaan Republik Indonesia (UKRI). The system is designed by integrating the Gemini Multimodal API with the Retrieval-Augmented Generation (RAG) framework to provide intelligent and efficient information services. The methodology includes building a campus-specific knowledge base from internal documents, which are processed into vector embeddings and stored in a Qdrant vector database. The RAG framework ensures that every response generated by the AI model is grounded in factual data from this knowledge base, thereby significantly improving accuracy. The implementation results show that the kiosk prototype successfully provides a natural and responsive multimodal (voice and text) interaction. Functional testing proves that the RAG mechanism dramatically outperforms a generic AI model in delivering relevant and accurate answers regarding campus information. In conclusion, this interactive kiosk has great potential to transform conventional information dissemination methods, making them more accessible and user-centric.