Ilm Nova AI: A Privacy-First Adaptive Learning Platform Integrating Six Evidence-Based Cognitive Science Methodologies

Adaptive learning platforms promise personalized education, yet most commercial systems operate as closed-source, cloud-dependent services that require student data to leave institutional premises. This paper presents Ilm Nova AI, an open, privacy-first adaptive learning platform that integrates six evidence-based cognitive science methodologies — Bayesian Knowledge Tracing (BKT), Zone of Proximal Development (ZPD), Bloom’s Taxonomy, SM-2 Spaced Repetition, Mastery Learning, and Retrieval Practice — into a single, atomically-updated engine. The platform runs entirely on local infrastructure with AI tutoring powered by locally-hosted large language models through Retrieval-Augmented Generation (RAG). All six methodologies update simultaneously after each learning event, enabling real-time difficulty calibration, spaced review scheduling, and mastery progression without transmitting student data to external servers.