Knowledge Explorer: An Agentic AI Framework for Interactive, Personalized and Multilingual Learning Experience

Abstract

Digital education tools have made learning more accessible, yet they often fall short in providing personalized guidance, multilingual support, and engaging pedagogical strategies. As global learners seek more adaptive and inclusive educational experiences, there is a growing need for systems that replicate the explanatory depth of expert tutors. Existing solutions typically lack fine-grained topic breakdowns, contextual grounding, and dynamic delivery methods. To address these gaps, we present Knowledge Explorer, a multi-agent AI system that delivers personalized, multilingual explanations using structured topic decomposition, retrieval-augmented generation (RAG), and storytelling. The system leverages LangGraph and LangChain to coordinate agents for subtopic division, factual retrieval, and narrative-based teaching, integrating Cohere’s SOTA embedding, rerank and LLM models alongside ChromaDB as the vector store. Our key contributions include a language-agnostic architecture for education, a storytelling-based delivery agent, and a RAG pipeline grounded in authoritative sources. Preliminary results show effective content generation in English, Hindi, and Spanish, underscoring the system’s potential as a scalable and globally adaptable educational companion. Overall, these capabilities position Knowledge Explorer as a promising tool for advancing learning equity, particularly in underserved and linguistically diverse communities.