Projects
2026
Why Does TikTok Know What You Like?
Interactive 15-minute lesson explaining recommendation systems through user signals, content clues, video ranking, and feedback. Includes worked examples, pair activities, speaker notes, and a closing quiz.
Project
GitHub
Knowledge-Guided Data Cleaning
Research survey connecting knowledge acquisition, repair selection, reliability, and intelligent data preparation across 30 studies. Includes an interactive literature map and survey paper.
AccessRoute HK: Accessible Mobility in Hong Kong
Accessible first- and last-mile navigation for Hong Kong, combining floor-level maps, mobility-aware routing, visual localization, obstacle reporting, and Cantonese guidance. Finalist and Gold Prize (Winner) in Track 1: Green Transport and Smart Mobility.
AI-Simulated ICU Multi-Agent System
Prototype ICU digital ward simulation with event-driven patient states, specialized agents, backend APIs, and frontend visualization.
2025
Elasticsearch and Madison Interactive Kibana Map
Indexed Madison geospatial, news, Wikipedia, and event data in Elasticsearch, answered search and aggregation questions, and visualized results through an interactive Kibana map.
ELT Data Pipeline with Airbyte, Snowflake, and dbt
Built a modern ELT pipeline using Airbyte for extraction/loading, Snowflake for warehousing, and dbt for transformations across survey, stock, foreign exchange, and CSV trading datasets.
Time Series Forecasting Kaggle Competition
Built forecasting models for a CS639 Kaggle competition, exploring EDA, feature engineering, EMA/ARIMA baselines, and gradient boosted tree models. Ranked first on the competition leaderboard.
LLM Fine-Tuning and RAG Course Chatbot
Explored quantized LLM inference, LoRA fine-tuning on lecture transcripts, and a RAG exam-preparation chatbot using Elasticsearch, Haystack, HuggingFace, and Streamlit.
Dynamic Factor Rotation for Quantitative Investment
Quant workflow for SSE 50 factor construction, regime identification, time-series prediction, dynamic weighting, and backtesting.
Predicting Citation Impact from Peer Review Data
PeerRead-based data analysis project studying whether review scores and metadata can predict future citation impact.
BadgerCS Career Nexus
Responsive web platform for UW-Madison CS students to browse jobs, view job details, use a forum, and prototype dashboards.
2024
Predicting Weibo Reposts, Comments, and Likes
Machine learning course project for predicting Sina Weibo engagement after publication, combining content, temporal, and user-profile features with Random Forest, XGBoost, SVM, RNN, and BERT experiments.
Quadratic Programming Solver Benchmarking
Numerical optimization course project implementing and benchmarking QP solvers, including IRWA and ADAL, against open-source optimizers across speed, accuracy, constraint violation, and stability metrics.
2023
Named Entity Recognition for FinTech
Financial NER project using LLM-assisted annotation, HMM implementation, CRF comparison, and evaluation visualizations.
Life Trajectory Classification Research
Fine-grained classification of life trajectories from Wikipedia using large language models and syntactic graph fusion. Accepted as an oral presentation at ICSC 2026.