The paper suggests a novel process-Adaptive Learning Style Inference with reinforcement-enriched ontology mapping (ALSIREOM), that is to drive the AI assistive-powered personalization of Immersion blended learning environments. The provided model dynamically varies the teaching material and the manner of interactions with the learners upon in-real-time inference of the need of cognitive, social and teaching presence of the learners along with the modes of preferred learning styles. The ALSIREOM system leverages fuzzy logic to model learners’ styles—drawing on frameworks such as VARK and Felder-Silverman—and integrates this with an ontology-based content classification framework that is continuously trained using reinforcement learning. The key contributions of the study include the following (1) it is practical to come up with an AI-powered personalization engine and learn the dynamics of learners preferences, (2) to maintain a healthy balance between the virtual and real worlds, Concepts of CoI need to be integrated and (3) the case study of undergraduate studies involving learners of STEM subjects demonstrates that the proposed personalization engine is feasible in aspects of enjoyment and performance.