TeachPro: A Unified Desktop Application for Intelligent Classroom Management and LLM-Powered Question Generation

Abstract - Teachers in contemporary institutions handle attendance, assessment, question preparation, and student data through disconnected tools, which increases administrative workload and reduces instructional focus. In this work, we developed TeachPro, a cross-platform desktop application built with MongoDB, Express.js, React.js, Node.js, and Electron.js that unifies eight teaching workflows and integrates a large language model pipeline (Google Gemini 1.5 Pro) for question and quiz generation. The system includes student management, notes, question bank, rich-text exam paper authoring with AI insertion, shareable AI quiz delivery, attendance analytics, timetable management, and secure JWT-based access. We present the architecture, implementation details, prompt-engineering strategy, and an evaluation protocol combining performance profiling and teacher pilot feedback. Across 100 generated questions, factual accuracy reached 84–96% across difficulty levels, Bloom’s taxonomy alignment reached 81–89%, and language clarity remained above 95%. The platform was built and tested in a real academic environment with faculty users during routine assessment preparation. Pilot use indicated 60– 70% reduction in repetitive administrative effort and faster end-to-end quiz creation. This manuscript provides a practical, reproducible blueprint for applying generative AI in everyday classroom operations with teacher-in-the-loop control. Key Words: AI in Education, Large Language Models, Gemini AI, Question Generation, Bloom’s Taxonomy, Electron.js, Teacher Productivity, EdTech, Attendance Management