As the digital education strategy continues to advance, traditional practical training in higher education faces challenges such as heavy workloads for instructors and delayed feedback when addressing large-scale, personalized demands. This study designs and implements an integrated training, assessment, and evaluation intelligent agent system tailored to practical training scenarios in science and engineering disciplines at universities, aiming to drive the transformation of teaching models from experience-driven to intelligent through artificial intelligence technology. The system’s core utilizes Alibaba Cloud’s Tongyi Qwen – Max large language model, integrating retrieval-enhanced generation technology to deeply empower a local domain-specific knowledge base. It has established a comprehensive functional framework covering the instructor, student, and administrative sides. The instructor side supports knowledge-base-driven intelligent lesson preparation, diverse question generation, and automated grading; the student side provides real-time online Q&A, personalized assessments, and error analysis. Developed using a B/S architecture with SpringBoot, MyBatis-Plus, and Vue 3, the system ensures a modular design with high cohesion and low coupling, as well as robust data security. Application results demonstrate that this intelligent system effectively reduces the burden of repetitive teaching tasks for instructors, provides students with precise and immediate learning support, and establishes a closed-loop process encompassing practical training, assessment, and evaluation, thereby significantly enhancing the intelligence and overall efficiency of practical training instruction.
