NearMissIQ

NearMissIQ is a web app (with optional desktop uploader) that helps autonomous vehicle teams find, triage, and learn from near-miss events—without waiting for crashes. It ingests vehicle logs (CAN, ROS bag, camera metadata, localization, planner outputs) and automatically flags “high-risk but non-crash” moments using configurable rules plus AI scoring. The product then clusters similar events, highlights the most common failure patterns (e.g., cut-ins, occlusions, phantom braking), and generates a reproducible replay package engineers can load into their existing simulation stack. The goal is to reduce time wasted combing through terabytes of data and to create a measurable feedback loop from road testing to model/planner fixes. Realistically, this won’t replace your internal tooling; it wins by being a focused layer for near-miss discovery and prioritization that plugs into what you already run.