Responsibilities include:

• Conduct research in recommender systems, social network analysis, and interdisciplinary data mining, with applications in healthcare, sports analytics, online platforms, and intelligent decision-making
• Teach courses in Data Mining and Graph Mining and Learning

Responsibilities include:

• Designed a power-efficient sleep stage classifier for wearable devices by bypassing 54.5% EEG signals to low-powered classifiers with a novel confidence-based switch, which reduces 32.7%-77.8% FLOPS with insignificant 1.9% accuracy drops (IEEE BigData 2024)
• Developed a multi-entity healthcare instruction recommender system for elders, incorporating unique demands of elders, diversities of individual multidisciplinary doctors, and their consensus-building process, resulting in 5-10% improvement in NDCG on synthetic data generated by domain experts (UHIMA&TLCMA 2025)
• Delivered in-person lectures on supervised learning, recommender systems, and real-time bidding as part of the Data Mining course (2025 Spring), totaling 9 hours

Responsibilities include:

• Designed a novel framework by recognizing and using the co-evolution on live stream platforms to recommend next-topics for live streamers, which outperforms conventional video recommenders by 27.1% in terms of precision and recall (CIKM 2023)

Responsibilities include:

• Leveraged viewer-streamer interactions on live streams and proposed a tensor co-factorization recommendation system, which reduces nearly 50% model parameters compared to SOTA methods (CIKM 2020)
• Identified a novel scenario of Virtual Reality (VR) group shopping and designed an effective recommender system and a query system, respectively (CIKM 2019 and VLDB 2020)
• Cooperated with experts from different domains, including Sociology (6 teams), Physics and Biology (15 teams from 4 research institutes), and developed machine learning models for social sentiment analysis and drosophila neuron identifications (Physics Reports 2023)
• Organized and analyzed results of four user studies and managed large-scale crowd-sourcing data annotations in four papers (one published in AAAI 2018, two in CIKM 2019, and one in VLDB 2020)

Responsibilities include:

• Used Lasso Regression model to improve the efficiency of CTR prediction by 61% without loss of precision

Responsibilities include:

• Incorporated a C4.5 Decision Tree model to identify causes of memory failure from sensor data with at least 98% accuracy and 40x speedup
• Built a rule-based FP tree model to generate reports of defects on wafers from manufacture data without human involving