Case Study: Zendar accelerates radar data curation and annotation with NomadicML

How Zendar Curates Data for Next-Gen ADAS Radar Models

Zendar, a developer of AI-first imaging radar systems for ADAS and autonomous driving, faced significant bottlenecks in training its perception models. The manual process of aligning and annotating radar and camera data was extremely slow, taking weeks to label just a few hours of footage. This led to inconsistent labeling quality and severely limited their ability to experiment and iterate quickly. To overcome these challenges, Zendar partnered with Nomadic and implemented its AI-driven data curation and annotation platform.

The Nomadic platform automated Zendar's data pipeline, using vision-language agents to annotate thousands of driving scenes for key events with human-like understanding. This solution integrated seamlessly with Zendar's cloud storage and SDK, enabling a continuous data flywheel. The impact was substantial: Nomadic helped cut data preparation time from weeks down to hours, drastically accelerated experimentation cycles, and ultimately improved model accuracy with higher-quality, standardized datasets.

Zendar

Antonio Puglielli

Vice President of Engineering