LLM-Enhanced Learning Environments for CS: Exploring Data Structures and Algorithms with Gurukul

In this Innovative Practice full paper, we introduce Gurukul, an innovative coding platform designed to support teaching Data Structures and Algorithm (DSA) course by integrating advanced Large Language Models (LLMs). LLMs have emerged as powerful tools in Computer Science Education (CSEd), offering unparalleled opportunities for enhancing student comprehension and engagement. However, their use in educational settings presents challenges, including tendencies toward hallucination, contextual inaccuracies, and the risk of undermining critical thinking by providing explicit solutions. To address these challenges, and to explore how specialized LLMs can bolster learner engagement, we present Gurukul, a platform featuring dual innovations: Retrieval-Augmented Generation (RAG) and Guardrails. Gurukul offers a hands-on practice feature where students can solve DSA problems within a code editor, supported by a dynamically Guardrailed LLM that prevents the delivery of explicit solutions. Additionally, the platform's study feature utilizes RAG, drawing from OpenDSA as a trusted source, to ensure accurate and contextually relevant information is provided. To assess the platform's effectiveness, we conducted a User Study with students, and a User Expert Review with faculty from a U.S. public state university specializing in DSA courses. Our analysis of student usage patterns and perceptions, along with insights from instructors, reveal that Gurukul positively impacted student engagement and learning in DSA, demonstrating the potential of specialized LLMs to enhance educational outcomes in this field.