Work

Projects that find the psychological structure in qualitative customer data and connect the evidence to strategy.

Probato

Independent research and prototyping since 2024. It represents customer reactions as stance, meaning, and Moral Foundations vectors, turning qualitative signals into evidence you can interpret.

Moral Foundations activation in customer reactions over time

Three-hour means of model outputs from a real case

Signals fell after the first May 18 interval. Fairness then stayed the highest signal in most three-hour buckets, at roughly 0.46-0.54. Vertical markers show major response events.

Sentiment distribution in online stakeholder reactions

LLM-assisted aggregate of 8,265 analyzable online reactions

The distribution is 6,009 negative reactions (72.70%), 1,044 neutral (12.63%), 717 mixed (8.68%), and 495 positive (5.99%). The sentiment labels are model outputs, not human ground truth.

SNU MBA Applied Business Project

An MBA graduation project that looks at what markets and stakeholders actually reacted to during a cybersecurity crisis, then weighs the response options against the evidence.

Trade Intelligence

A trade data dashboard commissioned by a Korean seafood importer. Public data from KCS and MFDS is unified into one screen with market signals, risk monitoring, import trends, and LLM answers grounded in verified data sources.

Market overview dashboard: trade KPIs, origin comparison, item seasonality, industry treemap

/dashboard/overview. Aggregate public-data view with no importer or manufacturer identifying information.

The overview screen shows trade volume, balance, import and export KPIs, origin comparison, item-level seasonality heatmap, and an industry treemap. Only aggregate counts are displayed; no individual company names are exposed.