First reported Sep 20 — we wrote this up later than the original.
General Robotics' GRID Platform Automates Full Robot Deployment Lifecycle
The agentic system cuts robot onboarding from weeks to hours and lets machines learn from every deployment.
US-based General Robotics has announced that its robotic intelligence platform, GRID, can now automate the entire lifecycle of robot development and deployment, according to a report by Robotics and Automation News picked up by South Korea's Robot Newspaper (로봇신문).
General Robotics describes itself as a company building an intelligence layer for physical AI. GRID was designed from the outset as an agentic system that continuously improves every robot connected to it. The company says GRID now automatically engineers its own processes, spanning everything from robot onboarding to skill deployment, which has sharply reduced the time and specialized robotics expertise needed to bring robots into real-world operation.
Ashish Kapoor, founder and CEO of General Robotics, said the industry has deployed advanced robotic systems the same way for decades. "For 50 years, people have deployed advanced robotic systems the same way. With AI advancing rapidly, it's time to rethink that approach," he said. He added that GRID's architecture unifies decades of accumulated robotics knowledge — research, software, AI models, simulation, data, and hardware — into a single agent-first platform.
At the core of GRID is a set of knowledge graphs that turn every deployment into structured, reusable intelligence. Each time a robot is onboarded, completes a task, adopts a new model, or fails at something, that experience feeds back into the platform, raising the baseline for all future work while keeping customer data and IP protected.
According to General Robotics, GRID's latest evolution has produced significant gains across four scenarios:
- Robot onboarding: from 1 month down to as little as 2 hours
- Model onboarding: from 3 days down to as little as 20 minutes
- Skill transfer across form factors: from 3 days down to as little as 1.5 hours
- New skill creation and deployment: achievable in as little as 2 days
The company says its customer base spans top-five global companies in automotive manufacturing, port operations, energy generation, and food and beverage production, along with several government agencies. These customers can freely choose whichever robot manufacturer or form factor best fits a given task.
GRID is also central to a growing ecosystem of robot makers offering transferable intelligence for rapid deployment and expanded capability, including FANUC, a leader in heavy industrial robots, and Galaxea, known for bimanual mobile manipulation. General Robotics counts Construct Capital, Accenture Ventures, E14, Nvidia, Shorooq, Valo Ventures, and Khosla Ventures among its backers.
A new way to build robots
General Robotics frames the robotics industry as constrained by two persistent problems: a shortage of engineers capable of designing intelligent robotic systems, most of whom are concentrated in research, and a highly fragmented development stack that forces custom integration of software tools, communication protocols, and programming paradigms for every different robot solution. Together, these constraints have made it difficult for companies to move beyond promising pilots into reliable, large-scale production systems.
GRID addresses this bottleneck with a modular library of composable skills, foundation models, and classical techniques that any connected robot can call on to complete a task. As underlying models and methods improve, GRID evolves alongside them, allowing robotic intelligence to be applied across different robot types and environments rather than being locked to a single AI approach or vendor.
An agentic system that engineers itself
With this latest evolution, GRID's architecture has been optimized for agent development and now automates the end-to-end robot engineering lifecycle. Given a task, GRID determines which skills are needed, selects the best combination of models and approaches, and automatically builds and runs the simulation environments required to train, evaluate, and refine the AI.
GRID also manages deployment of AI skills, monitors performance and errors, and determines what should change next. This closed loop — from onboarding through skill deployment, real-world evaluation, and feedback — runs continuously, with GRID constantly reasoning about what to build, test, fix, and improve next across every robot and task connected to the platform.
corrections · reports
Found a mistake? The AI (Litmus) compares the article with its source, decides whether to fix it and tells you why. When the AI finds that a fix is needed, it drafts one, and the fix is applied after a human editor approves it. Every fix is listed here and in the changelog.