Robotics start-up Skild AI on Tuesday unveiled “Skild Brain,” a foundation artificial-intelligence model designed to run on almost any robot, from assembly-line machines to humanoids. In demonstration videos the system enabled robots to climb stairs, stay upright when shoved and retrieve scattered objects—tasks that require real-time spatial reasoning and adaptive motor control. The Pittsburgh-based company said Skild Brain is trained on simulated episodes and human-action videos, then fine-tuned with data from each deployed robot. Early customers include LG CNS, while unnamed partners in logistics and other industrial sectors are testing the software. Robots in the field feed performance data back to Skild AI to continuously improve the shared model. Skild AI raised $300 million in a Series A round last year that valued the two-year-old firm at about $1.5 billion. Backers include Amazon.com, SoftBank Group, Lightspeed Venture Partners, Menlo Ventures and Sequoia Capital. Co-founders Deepak Pathak, the chief executive officer, and Abhinav Gupta say the goal is an “omni-bodied” brain that gives robots general intelligence similar to large language models but for the physical world.