The global impact of COVID in 2020 is forcing higher education institutions in many countries to adopt a hybrid model of education. The shift to blended education has been trending globally. However, the change in pedagogy has reduced the quality of students' learning experience and prevented them from deep learning. Therefore, the current study constructs a theoretical model using academic buoyancy and learner interactions as mediators, teaching presence, social presence, and cognitive presence as independent variables, and deep learning as the dependent variable, and proposes innovative strategies to promote deep learning in blended learning environments. The theoretical framework was empirically substantiated utilizing SPSS 26.0 and SmartPLS 4.0. The validation process employed the partial least squares (PLS) method within structural equation modeling (SEM) to scrutinize both the measurement and structural models. The findings demonstrated that (a) academic buoyancy functioned as a significant mediator influencing the relationship between teaching presence, social presence, cognitive presence, and the realization of deep learning; (b) learner interactions further emerged as an intermediary mechanism in the connection between teaching presence, social presence, cognitive presence, and the enhancement of deep learning; and (c) academic buoyancy was found to operate as a sequential mediator, impacting the relationship between teaching presence, social presence, cognitive presence, and deep learning via learner interactions. This investigation thus addresses a gap in the field by presenting a validated theoretical model for deep learning within blended learning contexts. It offers a fact-supported groundwork for optimizing blended teaching techniques, gauging the quality of learning, and developing actionable enhancement approaches.