Radareye offers a software‑only foundation model that fuses 4D imaging radar with camera data to provide reliable perception in rain, fog, snow, glare and other adverse weather conditions. The self‑supervised architecture learns from every driven mile without labeled data, runs on any 4D radar, and can be licensed as IP to existing vehicle silicon, enabling autonomous vehicle manufacturers and defense contractors to expand operational envelopes at near‑zero marginal cost.
Funding
Funding not disclosed
Founders
Product
Problem
Autonomous driving systems that rely primarily on cameras and LiDAR lose perception capability in rain, fog, snow, glare, and other adverse weather conditions, limiting their operational envelope and increasing regulatory liability.
Solution
Radareye provides a software‑only foundation model that fuses data from any 4D imaging radar with camera inputs to deliver robust perception in all weather. The model is trained in a self‑supervised manner, using every driven mile as unlabeled training data, which enables continuous improvement at near‑zero marginal cost. By being sensor‑agnostic, the solution can be integrated with existing radar hardware on autonomous vehicle platforms without requiring new sensors or hardware redesign. The approach aims to reduce disengagements where vision‑only stacks fail and where LiDAR performance degrades, thereby expanding the addressable market and mitigating liability in adverse‑weather scenarios.
Target Audience
Primary customers are autonomous vehicle manufacturers and defense contractors that need reliable perception capabilities across diverse weather conditions.
Features
- Fusion architecture that combines 4D imaging radar and camera data to maintain perception in rain, fog, snow, and glare
- Self‑supervised learning pipeline that leverages all driving miles as training data, eliminating the need for manually labeled datasets
- Sensor‑agnostic design that runs on any 4D imaging radar, avoiding hardware lock‑in
- Software‑only deployment model delivered as an IP license, compatible with existing vehicle silicon and requiring no new hardware
- Near‑zero marginal cost scaling, as model improvements are driven by accumulated driving data rather than additional sensor investments