This project aims to advance student engagement and achievement in high-enrollment STEM courses by substantively improving and evaluating CourseMIRROR, a mobile learning environment that delivers real-time, AI-guided reflection support. Powered by state-of-the-art natural language processing, CourseMIRROR prompts learners to reflect on what interests or confuses them, provides immediate feedback that spurs deeper thinking, and compiles class-wide insights for instructors. Partnering with universities and community colleges, the project reaches hundreds of students each semester and equips faculty with scalable, evidence-based practices that require no extra grading. By expanding access to effective study strategies and informing national priorities in AI-enabled education, the goal is to have broad benefits for retention and workforce readiness in science and engineering. Guided by the Reflection-Informed Learning and Instruction model and a Self-Regulated Learning (SRL) theory of change, the research pursues three integrated aims. First, adaptive prompts, motivational nudges, and automated reflection summaries are engineered and optimized through iterative usability and feasibility tests. Second, the effects of these features on motivation, emotion, SRL processes, and course performance are explored through field experiments across multiple introductory courses at multiple institutions. Third, multimodal data, including Motivated Strategies for Learning Questionnaire subs