Problem
- Black ice is hard to see and dangerous, and Busan has many icing-prone bridges, overpasses and tunnels
- Road management was reactive, with no system to warn about icing in advance
University & Competitions
Black Ice Risk Prediction Service
Predicts black ice risk for each district in Busan from weather data and serves it on a live map
Innovation Award · Popularity Award, 2025 BDIA Side Project
Period | 2025.10 ~ 2025.12
Invisible icing risk and road management that only reacts afterwards
My Role
- Backend: built a two-tier backend with FastAPI for model inference and Spring Boot as the gateway
- Implemented model loading, inference and an hourly batch schedule
- Applied admin overrides with 2-hour priority, separated public and admin APIs, and added JWT admin auth
Result
- A map dashboard updating risk levels (0-2) for 16 districts every hour from 14 KMA ASOS/AWS stations
- Predictions and admin edits stored in MariaDB for log-based reporting
- Won the Innovation Award and Popularity Award at the 2025 BDIA Side Project
