ko-br works on physical AI and embodied AI for manufacturing. The company's premise is that robots can be made to perform mundane manufacturing work reliably, provided they are trained on data gathered from real factory tasks. To that end it deploys robots in factory settings and collects data through human-in-the-loop teleoperation, with human operators guiding robots as they do genuine industrial work.
The data is sold to customers today as physical and embodied AI training data. ko-br states that its approach reduces factory operating expenditure by 50% from day one. Technical domains covered by the work include robotics, data collection, manufacturing automation and foundation models, with applications across manufacturing, factory automation and industrial robotics.
The company operates on a two-stage roadmap: selling collected data in the near term, and building in-house foundation models from that data so that robots can ultimately perform manufacturing work autonomously. Throughout, the emphasis is on reliability in real-world factory environments.
