CircuitAI: an Interactive Hardware Learning Platform

Abstract

Engineering students have been facing significant challenges bridging theoretical circuit knowledge with its practical implementation. While current general purpose AI systems exist, they have showcased inadequate accuracy in addressing hardware design questions, frequently providing incorrect instructions, flawed circuit analysis, and misleading debugging guidance, leaving students with no reliable assistance once outside of the classroom. This is where our project comes in: an interactive cross-platform hardware learning application that integrates three foundational components into a single system. The platform includes a Moodle-like learning management layer responsible for course structuring, enrollment, assignments, quizzes, grading, and progress tracking. It also provides interactive circuit simulation, featuring a real-time, client-side digital logic simulator for immediate feedback and conceptual validation, alongside a SPICE-based electrical circuit simulator capable of DC operating point analysis, DC sweep, AC frequency analysis, and transient time-domain simulation. Last but not least, a domain-specific AI engine, enhanced through Retrieval-Augmented Generation (RAG) and grounded in An-Najah University course materials and circuit design principles, which ensures that it delivers accurate, context-specific tutoring, and then we applied the same advanced AI capabilities to quiz generation and AI-assisted grading designed to support instructors while preserving human oversight. The resulting system enables students to learn theory, design circuits, perform simulations, and complete assignments within a unified workflow aligned with academic and laboratory practices. Integrating learning management, accurate circuit simulation, and curriculum-grounded AI ensures coherence between instruction, practical application, and evaluation. The front-end development was done using NextJS and CSS modules, while the back-end development was done in python for the AI inference server and setup, while the RESTful API endpoints were tied closely to our front-end via NextJS API routes. As for database choice, we used a PostgreSQL database for user data and general storage, and ChromaDB vector database for AI embeddings. While general circuit tools such as Multisim and Proteus support heavy simulation, and TinkerCAD and CircuitLab offer lighter educational interfaces, and while general-purpose AI assistants are becoming widely available, no project combines the best of all worlds into a single, easily accessible platform. This work demonstrates the effectiveness of such integration, with An- Najah University as its starting point.