Projects
A selection of things I have built — AI research, data pipelines, web apps and tools. Click a card for the details, or jump straight to the code.
- 2025
PitchMath
A football analytics platform. A pipeline pulls fixtures and stats from API-Football into Postgres, and a Streamlit app turns them into match, team and player insight.
- 2024
AIFORUS
A production AI-news intelligence platform I built solo at Aiforus (live at aicerti.co.kr). A Python pipeline continuously discovers, fetches and normalizes global AI-related news; a React dashboard maps and charts it. A public demo + its source are open.
- 2025
SafeStep
Accessible pedestrian routing for Seoul. Fuses OpenStreetMap walkways, public accessibility facilities (elevators, lifts, accessible toilets) and AI-detected street hazards, routed with a custom Valhalla engine behind a FastAPI web app.
- 2025
EvalKit
A unified CLI that benchmarks five AI model types — image classification, object detection, segmentation, speech and text — reporting not just accuracy but stability (metric variance across repeated data slices).
- 2024
Seaweed GAN
A DCGAN that generates synthetic dried-seaweed (gim) sheet images — a generator turns noise into seaweed textures while a discriminator learns to tell real from fake.
- 2024
Rice Leaf Classifier
A CNN image classifier (ResNet18) that identifies rice-leaf diseases across multiple classes from a Kaggle dataset.
- 2024
Insomnia
An automated pipeline that extracts sleep data from tracking-app screenshots (e.g. Samsung Health) and turns it into structured CSVs for insomnia analysis.
- 2025
ToolDock
A dashboard that launches a set of small utilities — audio, CSV analyser, hashing, image analysis and a system-info app. Built as Electron mini-apps; the browser-based ones run live on the web.
- 2025
NoticeHub
A unified viewer for Korean public-institution announcements (공지사항). Scheduled scrapers feed a Supabase notices table and a React + Tailwind UI lets you browse 245+ sources by region — deployed live on GitHub Pages.
- 2024
SeoulAir
Cleans, fills and merges scattered Seoul air-quality + weather exports (2022–2023) into analysis-ready datasets, then analyzes how the pollutants (PM2.5/PM10/O3/NO2/CO/SO2) relate to each other and the weather.