Projects
What I'm building and what I've built.
Currently working on
-
Message Board Sentiment and Retail Order Flow
In progressJune 2026 – Present · Independent two-person project
A study of whether the mood on Korean stock message boards matches what retail investors actually buy and sell. So far it has produced 1.37M collected posts and a fine-tuned sentiment model to label them.
- Python
- PyTorch
- Hugging Face
- KLUE-RoBERTa
- pandas
- statsmodels
Past projects
-
May 2026 – September 2026 · Personal research project
A research tool that predicts outcomes in Korea's sports-toto markets with rating and machine learning models, and a record of what its backtests do and do not show.
- Python
- scikit-learn
- MCP
-
April 2026 · Snowflake Hackathon, two-person team
A Streamlit app on Snowflake that recommends safer jeonse rentals across 288 Seoul neighborhoods, built by a two-person team at a Snowflake hackathon.
- Python
- SQL
- Snowflake (Snowpark)
- Streamlit
- pandas
-
December 2025 – March 2026 · Two-person project
A web app that collects Korean news articles for a keyword, groups articles about the same story, and writes a short summary for each group.
- Python
- FastAPI
- Next.js
- OpenAI API
- scikit-learn
- Vercel
