Designing a Course-Grounded AI Tutor with Retrieval-Augmented Generation: A DSR Approach to Technical Education

Intelligent tutoring systems (ITS) leveraging generative artificial intelligence (GAI) represent an emerging opportunity in educational technology, yet there is limited prescriptive knowledge on designing such systems for technical subject learning. In this paper, we conducted a Design Science Research project and propose design principles (DPs) for developing GAI-powered ITS that address current limitations in personalized learning at scale. We instantiated the proposed design and developed SmartStudy-a retrieval-augmented generative AI system with multimodal capabilities for machine learning education. The artifact autonomously processes textbook content, generates practice tests, provides grounded answers, and performs automated grading. We evaluated SmartStudy with 44 data science students using performance metrics and user feedback assessments. Results demonstrate improved student performance, positive learning effects, and high continuance intention. This paper provides practitioners with design principles for creating GAIbased tutoring systems with concrete guidelines that can be extended to other technical domains.