On November 1, the closing ceremony and outstanding project roadshow of the Large Model Application Development Training Camp at Southeast University was successfully held in the university lecture hall. At the roadshow, students showcased 16 intelligent agent application achievements in groups and delivered on-site defenses, featuring brilliant presentations and enthusiastic interactions. The defense session was broadcast live online simultaneously, attracting over 6,000 concurrent viewers. In the end, a number of outstanding projects covering AI-enabled education, mental health & wellness, scientific research, entertainment, career development, travel and other fields stood out for their outstanding innovation, solid technical implementation and clear application potential. With this, the three-week training camp came to a successful conclusion.
Jointly launched by Southeast University, Inspur Information and the Datawhale Open Source Learning Community, the training camp represents an in-depth exploration of school-enterprise collaborative practical teaching, providing a replicable model for cultivating interdisciplinary talents under the "AI+X" framework. It attracted postgraduate and doctoral students from more than ten schools, including the School of Cyber Science and Engineering, School of Electronic Information Engineering, School of Software Engineering, School of Electronic Science and Engineering, and School of Architecture. During the three-week intensive training program, 154 students built large model applications from scratch on the MetaBrain Enterprise Intelligence EPAI Platform via a dual teaching model combining theoretical lectures and hands-on practice. They transformed abstract theories into executable workflows and turned interdisciplinary creative ideas into reproducible intelligent applications. Gradually, they achieved core capability upgrades ranging from basic AI literacy, intelligent agent workflow construction and RAG technology practice to large model fine-tuning, exploring the infinite possibilities of "AI+X".
Witnessing the Up surging Trend of AI Learning
From creative brainstorming to project delivery:
154 postgraduate & doctoral students from 10+ schools (80% postgraduates, 10% PhD candidates)
Students from the School of Cyber Science and Engineering formed the largest participating group
Over 40 students attended brainstorming sessions with active discussions and inspiring exchanges
60 creative proposals submitted, incubating more than 30 intelligent agent applications; 16 high-quality projects shortlisted for the Demo Day roadshow
Over 6,000 concurrent online viewers for the defense live stream
Outstanding Project Showcase: Exploring Integrated Innovation in Diverse AI+X Scenarios
Solid understanding enables earnest practice, and earnest practice ensures clear and resolute execution. Throughout the training camp, students applied theoretical knowledge to real-world scenarios. Based on the MetaBrain EPAI Platform, they completed full-process development including prompt engineering, knowledge base construction and intelligent agent workflow design, launching over 30 practical intelligent agent applications. Selected outstanding projects covering AI+ mental health & wellness, scientific research, entertainment and other fields are showcased below. These deliverables feature robust technical performance, smooth user interaction, and fully demonstrate students’ scenario insight and interdisciplinary innovation capabilities.
AI + Mental Health & Wellness
Chat Health integrates the dual roles of a personal smart fitness coach and a professional nutrition consultant. Powered by the Qwen3 large model as its decision-making core, it adopts an innovative workflow to precisely categorize and process fragmented health needs including diet guidance, exercise plans and popular science knowledge. The application can instantly calculate BMI and generate customized solutions based on users’ real-time height and weight data. It also stores personal health records via memory nodes and parameter extraction, continuously learning user profiles and historical data to provide dynamically adjusted, highly personalized health management plans.
Personalized Music Emotional Support combines intelligent dialogue with music therapy to build an emotional companion with in-depth emotional understanding and customized music recommendation capabilities. Supported by a multi-model collaborative workflow, it accurately identifies and analyzes negative emotions such as sadness, anxiety and stress. Based on music psychology principles, it generates dynamic music recommendations tailored to rhythm, harmony and style. With emotional dialogue, scenario-based music descriptions and 24/7 online companionship, it delivers an immersive emotional healing experience.
AI + Scientific Research
The Feynman Learning Assistant draws on the Nobel laureate Richard Feynman’s "learning by teaching" philosophy, integrated with AI technology to create an intelligent tutor that identifies knowledge gaps and guides in-depth thinking. Two refined prompt sets are designed: one analyzes user inputs to detect missing core concepts, logical flaws or improper terminology, and offers constructive feedback; the other generates easy-to-understand Feynman-style explanations with vivid analogies for reference. Adopting an interactive workflow of "user explanation - AI feedback - reflection & revision", the assistant replicates the core loop of the Feynman Learning Method, helping students identify blind spots and reinforce knowledge retention effectively.
SEU Library Guide is an efficient and reliable campus knowledge service assistant built on large language models and RAG technology. Its core innovation lies in multi-agent collaborative workflow design: a demand organizer refines user inquiries, a web content analyzer retrieves real-time official information, a knowledge synthesizer generates multiple candidate answers, and an answer evaluator outputs the optimal response through weighted scoring based on authenticity, practicability and accessibility. This mechanism greatly improves answer accuracy and practicality, solving common pain points of traditional library Q&A systems such as slow response, generalized content and inaccurate recommendations.
AI + Entertainment & Practical Services
The Intelligent Travel Planning Assistant provides one-stop travel services from natural language understanding to customized itinerary generation. It automatically identifies user intentions through semantic analysis and diverts requests to itinerary planning or knowledge Q&A modules. The former delivers structured reports with destination overviews, daily routes and practical tips; the latter offers precise answers based on a dedicated travel knowledge base. The assistant can automatically supplement key information such as budget and preferences from vague demands, showcasing AI’s value in improving travel efficiency and satisfaction through intelligent scheduling and scenario-based suggestions.
The Intelligent Job Application & Interview Assistant builds a full-spectrum career solution covering job profile analysis, quantitative resume evaluation, targeted optimization, intelligent job matching and mock interview generation. Technically, it adopts multi-role prompt simulation to replicate virtual experts including HR specialists, interviewers and resume consultants. It conducts in-depth analysis of target positions, calculates resume-job matching scores, and provides actionable revision advice following the STAR framework. Powered by data-driven intelligent assessment, it delivers precise and efficient employment support for graduates and young professionals.
Award Ceremony: Celebrating Moments of Glory
Sixteen student teams won major awards for their excellent technical strength, innovative ideas, practical application value and commercial potential.Swipe left or right to view more pictures.
Embracing AI in Practice: Exploring a New Model for AI+X Interdisciplinary Talent Cultivation
Chen Yang, Deputy Party Secretary & Vice Dean, National Outstanding Engineer College, Southeast University
This Large Model Application Development Training Camp is a vivid practice of integrating industry and education. The school-enterprise collaborative practical education model has successfully built a bridge from theoretical learning to innovative practice. Guided by real demands and driven by technological innovation, students developed intelligent assistants for academic tutoring, scientific research collaboration and career development. These creative achievements not only demonstrate the application potential of large model technologies in diverse scenarios, but also reflect students’ ability to integrate cutting-edge technologies with practical needs.
Professor Dong Fang, Doctoral Supervisor, School of Computer Science and Engineering, Southeast University
Enabling non-computer majors to master large model application development has become crucial for AI+X talent cultivation. Amid rapid technological iteration and interdisciplinary integration, AI is increasingly valued as a practical tool to solve cross-disciplinary problems, rather than abstract theories. Effective teaching intervention is essential to help students overcome initial fears and turn ideas into deliverables, which has become a key topic for university AI general education.
Student Ke
This has been an unforgettable training experience! The focused and collaborative atmosphere among teammates motivated me to polish our project with greater passion. This is just a starting point, not the end. We will keep improving our achievements from 1 to 10, to 100 and beyond. Only by staying open-minded and learning from each other can we break cognitive boundaries and pursue continuous progress. Learning is a lifelong journey.
Student He
I am deeply grateful to the university leaders, as well as engineers from Inspur Information and Datawhale. Teachers provided full support and hands-on guidance throughout the training, especially on platform operation and practical tasks. This camp completely changed my understanding of large models. I used to feel they were both familiar and distant: we use them in daily study and life, yet know nothing about their underlying mechanisms as a "technical black box". I will apply all the knowledge and skills I have learned to my future research and exploration.
Student Li
This training camp offered me a precious opportunity to turn abstract large model concepts into concrete instructions and practical applications, and we achieved rewarding results. Before participating, my understanding of large models was limited to basic dialogue and Q&A. The systematic courses opened the door to large model application development for me. Learning about intelligent agent workflows inspired me greatly, as I realized AI can be designed and assembled like an assembly line with modular thinking.
This training camp marks a new starting point and lays a solid foundation for future exploration of AI+X interdisciplinary talent cultivation. Moving forward, Inspur Information will further deepen school-enterprise practical education cooperation, joining hands with universities and industry partners to build a future-oriented AI+X talent ecosystem. We will continuously transform advanced engineering platforms into high-quality teaching resources, empowering more students to grow through integrated industry and education, and forge a replicable, sustainable path for cultivating outstanding AI+X interdisciplinary talents.
Reposted from the WeChat Official Account: MetaBrain Server
