The company calls the approach Auto-Engineering. Instead of requiring robotics engineers to manually handle every stage of calibration, simulation, model integration, skill development and deployment, GRID is designed to perform much of that work automatically and learn from previous deployments.
General Robotics says the update has reduced robot onboarding from roughly one month to as little as two hours, while model ingestion has fallen from around three days to as little as 20 minutes. Transferring an existing skill between different robot form factors can take as little as 1.5 hours, according to company-reported testing. These figures have not been independently benchmarked across the wider robotics industry.
GRID works across industrial arms, humanoids, quadrupeds, wheeled systems and drones. In one example described by the company, the platform learned a new robotic skill from a short video by reconstructing the task in simulation, generating training data and then deploying the resulting policy onto physical hardware.
The company says GRID is already deployed with approximately a dozen enterprise customers spanning manufacturing, logistics, energy and government. CEO Ashish Kapoor told GeekWire that General Robotics now generates revenue in the millions of dollars.
General Robotics was founded by former Microsoft robotics researchers and has raised nearly $34 million. Its investors include NVIDIA, Khosla Ventures, Construct Capital, Accenture Ventures and Valo Ventures. Accenture also integrated GRID into its broader physical AI strategy earlier this year.
The development targets an increasingly important bottleneck in robotics: not building more capable hardware, but reducing the engineering work required to turn new robots and AI models into systems that can reliably perform useful tasks in production.
