Encord, a data annotation company, is partnering with German neuroscience startup Zander Labs to measure brain waves from human "pilots" who wear sensor-equipped headsets while performing physical tasks like disassembling block towers. The goal is to capture mental states like error, intent, and surprise to create more useful training data for humanoid and warehouse robotics models.
The effort targets what industry insiders call the "bleeding edge" of solving the robotics data bottleneck. Encord's head of robot learning Vineeth Velmurugan, a veteran of OpenAI's robot lab and warehouse automation firm Berkshire Grey, says the company realized customers applying end-to-end learning to robotic manipulation would need to produce training data themselves rather than simply manage it.
The current project with Zander Labs is a trial run—Encord plans to build an initial brain wave-tagged data set, test it through customer robotics models, and evaluate whether it actually improves performance before scaling. Zander neuroscientist Lucas Gehrke notes that brain activity patterns during tasks can signal when models need to deploy their highest-effort processing.
Velmurugan estimates breaking through current limitations would require a dataset roughly five times the size of YouTube's video corpus, explaining why data generation has become a business rather than just a research problem.