Contact  Us

Nexar operates one of the largest real-world video driving datasets and turns it into AV training data, collision-prediction models, and real-time road intelligence for companies building autonomous vehicles, mapping, and smart-city systems.

It is captured from real roads, not generated. A network of 350,000+ connected dashcams records authentic, unstaged driving, including the rare long-tail events simulators miss, providing ground-truth data for training and validating AV and ADAS models.

Nexar's network captures over 1.2 billion miles of real driving each year from 350,000+ connected cameras, including more than 60 million long-tail edge-case videos.

BADAS is Nexar's collision-prediction model. BADAS 2.0 reaches 99.4% precision and 91% early-warning recall, trained on 2 million real-world clips, and outperforms a fine-tuned NVIDIA Cosmos model at 91x fewer parameters.

APEX benchmarks an autonomous-vehicle stack against billions of miles of ground-truth human driving data, giving a real-world performance measure instead of simulation-only validation.

The City AV Index scores each U.S. city using real-world driving behavior, monthly data freshness, and block-level risk signals.