AURA - AI POWERED ADAPTIVE LEARNING

Most online courses give every learner the same lessons at the same pace, so gaps in understanding often go unnoticed. This presents AURA (AI-Powered Adaptive Learning System), a web application that generates a course from a learner’s goal and adapts it to the learner’s quiz results. A large language model (LLM) creates a course outline with modules, lessons, topics and prerequisite links, and each lesson is written when the learner opens it. A tutor based on retrieval-augmented generation (RAG) answers questions from the current lesson and cites the passages it used. After each multiple-choice quiz, a weighted moving average updates a mastery score and a confidence value for the topic, and fixed rules select the next step and the next quiz difficulty. All LLM output is validated against schemas. In a pilot run with the gpt-5.4-mini model, five of seven generation jobs succeeded on the first attempt without repair. Two failed, one because of a provider quota error and one because the safety-check output was invalid. A course outline and first lesson were ready in 26.6 s. The prototype supports multiple-choice questions only, uses fixed adaptation rules and has not been tested with real learners.