Development of an Intelligent Chatbot for Streamlined Support of Non-EU Students Exploring Study Opportunities in Finland Utilizing a Large Language Model

Abstract

This thesis presents the development and implementation of an AI-powered chatbot specializing in Finland study and visa services for non-EU students. The project aims to create an efficient and user-friendly chatbot prototype to assist prospective international students navigating the complexities of studying in Finland. The main content of this thesis covers the entire development process of the chatbot. It details the use of language models, specifically the fine-tuning of LLaMA 3 8B Instruct model using Quantized Low-Rank Adaptation (QLoRA) techniques. The implementation incorporates a Retrieval-Augmented Generation (RAG) system to enhance the accuracy and relevance of responses. The thesis also explores the development of a user-friendly interface using Streamlit and the integration of Google OAuth 2.0 for secure user authentication. The outcome is a functional AI chatbot capable of providing accurate, context-aware responses to queries about studying in Finland, visa requirements, and related topics. The chatbot demonstrates the ability to maintain conversation context, handle multi-turn interactions, and provide up-to-date information by leveraging its RAG system. this project showcases the potential of AI-driven solutions in improving information accessibility for international students. It highlights the effectiveness of combining advanced language models with information retrieval systems to create a specialized, domain-specific chatbot.