Enhancing University Education with AI: A Telegram Bot Leveraging RAG and External APIs for Secure Knowledge Retrieval

This paper presents a novel AI-powered Telegram bot designed to enhance university information services by securely integrating external AI capabilities with institutional private data. The system leverages Retrieval-Augmented Generation (RAG) to transform structured university data (faculty profiles, schedules, lecture notes) into vectorized embeddings, which are dynamically retrieved and combined with responses from a general-purpose AI API (e.g., GPT-4). This hybrid approach ensures accurate, context-aware answers while preserving data privacy — raw institutional information is never exposed directly to third-party systems. Implemented at Comtrade University, the bot demonstrates significant outperforming standalone AI models for domain-specific questions. Key innovations include a scalable pipeline for embedding private data, seamless Telegram-based access, and cost-efficient prompt engineering via RAG. The solution addresses critical challenges in educational technology: balancing AI augmentation with data security and providing 24/7 conversational access to institutional knowledge. We discuss architectural decisions, privacy safeguards, and empirical results, offering a replicable framework for other universities.