General Robotics Says GRID Platform Automates Robot Engineering, Cutting Setup to Two Hours

The Redmond, Washington-based company sells what it calls an intelligence layer for physical artificial intelligence. GRID connects robots built by different manufacturers to a shared library of skills, foundation models and older control techniques that any machine on the platform can call on to carry out a task.

Other steps in the process have shrunk by comparable margins. Bringing a new AI model into the platform now takes as little as 20 minutes, down from three days. Moving a skill from one robot form factor to another has fallen from three days to a best case of 90 minutes. Building and deploying a skill that did not previously exist takes as little as two days.

Given a task, GRID determines which skills it needs, assembles a combination of models and methods to produce them, then builds and runs the simulation environments required to train and evaluate the result. Deployment follows, and the system then tracks performance and failures in the field to decide what to change next. That loop runs continuously across every robot connected to the platform.

Underneath sits a set of knowledge graphs. Each robot onboarded, task performed, model ingested and failure logged is stored as structured material that later deployments draw on, so each job is meant to leave the platform better at the next one. General Robotics said customer data and intellectual property remain walled off.

"People have been trying to deploy advanced robotic systems the same way for 50 years. With the rapid advances in AI, it's time to rethink the approach," Chief Executive Ashish Kapoor said. He described GRID as an agent-first platform that pulls research, software, AI models, simulation, data and hardware into one place.

Fanuc America Corp. and Chinese robot maker Galaxea Dynamics are both working with the platform. Ryan Patterson Jr., a general manager in Fanuc America's robotics division, said General Robotics is an authorized system integrator for Fanuc and is advancing frontier physical AI across its product line. Galaxea founder Gao Jiyang said the two companies are speeding up deployments in automotive, pharmaceutical and consumer packaged goods.

Customers include government agencies and companies that General Robotics describes as global top five in automotive manufacturing, port operations, power generation and food and beverage production.

Hans Peter Brondmo, former chief executive of Everyday Robots and a former vice president at Google X, said in the announcement that robotics has no single "ChatGPT moment" coming. Progress arrives deployment by deployment, he argued, built on real-world data that most companies are not set up to capture.

Kapoor founded the company in 2023 with fellow Microsoft Corp. researchers Sai Vemprala and Shuhang Chen, under the name Scaled Foundations, and renamed it last year. He spent 17 years at Microsoft, including running its autonomous systems and robotics research group. General Robotics is venture capital funded and counts among its investors Construct Capital, Accenture Ventures, Nvidia Corp., Khosla Ventures, E14 Fund, Shorooq Partners and Valo Ventures.

Editor's Summary

General Robotics says its GRID platform now carries out much of its own engineering work, cutting the time to get a new robot into production from about a month to as little as two hours and shortening related tasks such as ingesting AI models and porting skills between robot types. Fanuc America and Galaxea Dynamics are working with the platform, and the company lists government agencies and large manufacturers among its customers. The claim rests on the company's own account of the platform, with no independent verification of the timing figures.