Development and evaluation of AI-powered educational assistant

Abstract

In order to improve student learning in college courses, this thesis investigates the creation and application of J.A.R.V.I.S. (Just An Resourceful Virtual Instructor and Study-aid), an AIpowered instructional assistant. Through the use of sophisticated Natural Language Processing (NLP) methods, such as refining models like as LLaMA-3 and incorporating pre-made solutions like Livechat AI, J.A.R.V.I.S. offers accurate, context-sensitive support in response to student enquiries. The research is organised around two main strategies: the first is optimising the LLaMA-3 model using Alpaca and Unsloth to produce a highly specialised teaching assistant that can produce responses that are rich in context and information. The second method uses Livechat AI to incorporate the J.A.R.V.I.S. chatbot into a specially designed website, with scalability, userfriendliness, and real-time communication as top priorities.The system's performance was assessed using two main criteria: its feasibility in real-world deployment and its capacity to manage intricate educational issues. The expected result is a notable enhancement in the understanding and involvement of students, since J.A.R.V.I.S. effectively provides precise and pertinent answers customised to meet individual learning requirements. By showcasing the use of cutting-edge natural language processing techniques to improve learning outcomes and providing insights into the advantages and practical difficulties of implementing artificial intelligence in educational settings, this study advances the area of educational technology.