Fullstack and AI/LLM-powered systems, plus HCI research work shipped as production apps.
Python pipeline that diffs DART filings against the prior version and uses Gemini to summarize changes and flag risks, served to a React dashboard through a Spring Boot API.
Automated twice-daily DART filing analysis (06:00 / 18:00 KST)
LLM app that turns any public GitHub repo into documentation, interactive UML diagrams, and quizzes, cutting codebase ramp-up from hours to under 60 seconds.
Ramp-up time: hours → under 60 seconds
Personal-training marketplace with an end-to-end booking flow (calendars, reservations, session history) and full Toss Payments integration, including cancellations and refunds.
Full payment lifecycle: ready → success → cancel → refundAndroid front-end for a wearable that shows real-time oral pH readings from a color sensor. The work was co-authored into 2 peer-reviewed publications.
Co-authored 2 peer-reviewed publications (ACM, INTERACT 2025)Partnering with hardware, backend, and research teams — and writing the design/test documentation that turned research findings into behavior non-engineers could understand and use, across 2 peer-reviewed publications.
Building with Claude Code and Gemini as daily tools, not just APIs — pairing structured data with LLM reasoning instead of over-engineering brittle parsers, e.g. redesigning a filing-analysis pipeline to combine rule-based fields with LLM-powered semantic analysis.
Comfortable owning the full stack — relational schema design, payment lifecycles, and deployment — moving between Android/Kotlin, Java/Spring Boot, and React/Node.js as the problem demands.
Certified: Introduction to Model Context Protocol (Anthropic)
Interactive Organisms Lab — Dr. Katia Vega, UC Davis