Engineering education is currently undergoing a transformation driven by two converging forces: the rapid evolution of domain knowledge and the disruptive impact of Generative AI on learning practices. These changes place unprecedented demands on Learning Management Systems (LMS), which must increasingly support adaptive, data-driven, and personalized learning experiences. However, widely used commercial LMS platforms, such as Moodle, eLearning suites, or CourseGarden still offer only limited personalization capabilities and provide restricted access to fine-grained learner activity tracking data. This lack of transparency hinders both instructional innovation and research efforts aimed at advancing educational technologies.

To address these limitations, the Graph4Learn (G4L) Intelligent Tutoring System was developed. The system is built upon the Evolving Knowledge Space Graph (EKSG) model, a structured graph-based knowledge representation designed to support rapid curriculum evolution, personalized learning path generation, and knowledge state prediction. The G4L framework also incorporates mechanisms for handling the forgetting phenomenon, thereby enabling more accurate and realistic modeling of knowledge retention over time. In addition, G4L provides detailed learner activity logs, making it possible to implement and evaluate a wide range of knowledge tracing algorithms for research and instructional development purposes.

In this presentation, I introduce the core components of the Graph4Learn framework and demonstrate its functionality through a case study conducted with 45 university students enrolled in a Databases course. The G4L system is implemented as a Python-based web application, employing a RESTful peer-to-peer architecture to facilitate communication among its distributed components. This architectural choice significantly improves scalability and modularity, which are critical for effective adaptation to rapidly evolving technological landscapes. The EKSG model is realized within a Neo4j graph database, primarily selected for its inherent capability to efficiently represent and query the intricate relationships between knowledge concepts using Cypher queries. This graph-native approach offers a distinct advantage in managing the complex characteristic of a dynamic knowledge domain. The learner profiles and activity logs are systematically stored in a MySQL relational database, providing robust and efficient management of structured individual progress data. The specific learning content, tightly linked to the EKSG model, is presented in concise two-page PDF documents, optimizing accessibility and focus. Furthermore, quizzes are stored in YAML files, which ensures ease of authoring and readability for domain experts and instructors, thereby facilitating agile content updates essential in such dynamic fields.

During the study, students used the G4L system to learn a selected topic, navigating through graph-structured learning materials and receiving personalized recommendations based on their dynamically inferred knowledge state. The results indicate that students responded positively to the system, reporting improved clarity in understanding the structural relationships within the domain. Moreover, the collected data reveal that graph-based knowledge representations can significantly support learners in visualizing prerequisite relations, identifying appropriate next learning steps, and self-regulating their study process.

Overall, the Graph4Learn framework demonstrates how intelligent tutoring principles, combined with flexible graph-based models, can provide an effective response to the emerging challenges of engineering education. The system not only enhances the learning experience but also offers a robust platform for future educational research in adaptive learning, learner modeling, and AI-supported instruction.