Project
Project
A Multimodal Deep Knowledge Tracking Model and its Application to Intelligent Recommender Systems
Researcher 03/2023–8/2024
Participated in the "Guangdong Province Educational Science Planning Project (Higher Education Specialization)" and
iFlytek's project "Research and Practice on AI-Enabled Classroom Teaching Skills Training for Pre-Service Teachers."
Constructed and optimised a deep knowledge tracking network for conducting multimodal course tracking and analysis, achieving 1.12% and 1.75% improvements in AUC and ACC over baseline models, outperforming similar models.
Assembled and preprocessed a large educational data set by collecting student reflections and practical results for a five-semester period using Python.
Designed and implemented a recommendation system offering personalized learning resources based on model predictions, supporting student learning at different stages
An Automated Driving Fatigue Detection System Based on Status Analysis Algorithm
Researcher 06/2023–08/2023
Developed Efficientnet-B7 convolutional neural network using PS-KD algorithm to evaluate driver status.
Optimised evaluation accuracy by utilising test time augmentation (TTA) for data enrichment.
Accelerated formulation process for relative codes and models, achieving recognition accuracy of 99.96%.
Took responsibility for model code testing, including development of testing script and algorithm introduction.
An Automatic Application System of Abnormal Learning State Detection Based on Computer Vision
Project Manager 03/2023–08/2023
Presented improvements to facial expression recognition model on Real-world Affective Faces Database (RAF-DB), increasing recognition accuracy by 3% compared to other approaches.
Monitored and assisted with the development of an automatic monitoring system for abnormal detection using PyQt framework based on YOLOv5 and revised Self-Cure Net.
Oversaw development of target system, including formulating performance test reports and technical documents for functional introduction.
Publications
LDFacilitator: A system for facilitating learning design using large language models.
Tang Shan, Song Jiajia, Shen Jie, Wang Nan, Law Nancy W. Y., & Lin, Jionghao. (2025, May).
Interactive Event (Demo) paper presented at the 26th International Conference on Artificial Intelligence in Education.
MTFNet: Multi-Scale Transformer Framework for Robust Emotion Monitoring in Group Learning Settings
Yi Zhang* ; FangYuan Liu* ; JiaJia Song; Qi Zeng; Hui He
2024 Asia Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC)
Work Experience
Shenzhen Data Thinking Technology Co., Ltd.
Computer Vision(CV) Algorithm Engineer (Intern) 07/2022–08/2022
Captured and labelled 1,000+ single and multi-person image data sets according to task requirements.
Reinforced visual recognition capacity of mobile devices by applying revised YOLOv5 algorithm structure, achieving 5% accuracy improvement from the original model.
Contributed to the development of crowd detection models based on artificial intelligence computing platform.
Conducted in-depth research examining current mainstream PyTorch AI framework and formulated relevant research report based on testing results for various object recognition networks.
Tamme (Hong Kong) Limited
Developed an intelligent agent system based on LLM, constructing a real-time emoji and video generation pipeline that utilizes self-forcing and VACE to manage character expressions.
Utilized n8n to create an automated agent workflow, achieving complex task decomposition, resource scheduling, and dynamic collaboration strategies for the Agent.
Built and deployed a graph memory system for LLM Agents.