A new startup is addressing a major bottleneck in physical AI and autonomous systems by converting unused fleet video footage into structured, searchable datasets that can be used to train and improve AI models. Backed by TQ Ventures, Pear VC, and AI leader Jeff Dean, the company is building data infrastructure for autonomous technology.
The platform converts raw video from autonomous vehicles and robotics systems into AI-ready data, identifies rare edge cases, and uses vision-language models to make footage searchable and usable for training. Edge cases are critical in autonomous systems because they help AI models learn how to respond to unusual or high-risk real-world scenarios, which are often the hardest to capture and label.
As autonomous technology scales across self-driving cars, robotics, drones, and physical AI systems, the biggest challenge is no longer just building models, but finding, structuring, and training on high-quality real-world data. This is where AI data infrastructure platforms are becoming increasingly important.
In the autonomous AI race, data quality and edge-case training are becoming a major competitive advantage.
Bottom line: The startup is positioning itself as a key player in the physical AI data infrastructure layer, helping autonomous companies train smarter models using real-world video data.
