What it is
BicycleRadar is a predictive collision-warning system for cyclists — it watches what’s approaching from behind and warns the rider before a close pass becomes a crash. I built it after one too many close calls on my own weekend rides.
How it works
Sensor fusion plus a lightweight ML model running on the edge produced a 3.2-second warning window at 96% prediction accuracy — enough lead time to actually react. The hard part wasn’t detection; it was keeping false alarms low enough that riders trust it.
Why it matters
Safety systems only work if people believe them. Tuning for trust — not just raw accuracy — is the thread that carried into everything I built afterward.