Results come one year after launching out of stealth, showcasing the company’s key achievements in building its robotics foundation model
Skild AI, an AI robotics company building a scalable foundation model for robotics, today released a never-before-seen look at its results. The video highlights the company’s early progress in scaling the capabilities of Physical AI—its shared, general-purpose brain that’s designed for a diverse embodiment of robots, environments, and tasks.
This press release features multimedia. View the full release here: https://www.businesswire.com/news/home/20250729431330/en/
“Robotics is marred by Moravec's paradox: the hard problems are easy and the easy problems are hard. A lot of current robotics models focus on tasks that are hard for humans and easy for robots: dancing, kung-fu, because they are free-space actions and do not require any generalization,” said Deepak Pathak, CEO and Co-Founder of Skild AI. “Skild AI models can not only solve these easy tasks but also solve everyday hard tasks such as climbing stairs even under adversarial conditions, or assembling fine-grained items, which require vision and reasoning about contact dynamics.”
Abhinav Gupta, President and Co-Founder of Skild AI, said, “Unlike other groups that focus on a vertical approach (focus on a single robot, or single task), Skild AI showcases the general purpose nature of Skild’s Brain: the same brain can be adapted to different tasks and different robot morphologies. From quadrupeds to humanoids, from table-top arms to bimanual AMRs; from home tasks such as dishwashers to physically challenging tasks such as climbing slippery slopes—Skild Brain is a one-stop AI Model for all different robot forms.”
The released video also highlights that Skild Brain is robust and has enough generalization to co-exist with humans. Skild Brain is not only safe around humans (thanks to end-to-end AI models, which are trained to apply low forces), but it is also highly adaptive to disturbances and human interactions. This feature has often been limited or missing in other AI models.
The data puzzle
One of the biggest challenges in building a robotics foundation model is the lack of any large-scale robotics data, and collecting real-world data using hardware is slow and prohibitively expensive. Skild AI has tackled this challenge head-on: achieving scale by leveraging large-scale simulation and human videos on the internet to pre-train its foundation model at scale. Skild AI post-trains this foundation model with targeted real-world data to deliver working solutions to its customers.
Skild trains models that can work across different morphologies, including human data, vastly expanding the available training set. This has enabled Skild to build a truly physical foundation model, which can work on quadrupeds, humanoids, tabletop, and mobile manipulators. Trained on diverse data, tasks, and hardware, Skild’s general-purpose brain is fine-tuned and distilled to meet deployment needs.
The team
Skild AI’s team has been behind key robotics breakthroughs such as self-supervised learning, curiosity-driven exploration, large-scale imitation, sim2real locomotion, parkour, dexterous manipulation, learning from internet-scale video, and more. Team members have also built production-ready systems at Anduril, Tesla, Nvidia, Meta, Kitty Hawk, Google, Everyday Robotics, and Amazon.
Skild AI is backed by investors including Lightspeed Venture Partners, Coatue, SoftBank Group, Jeff Bezos (through Bezos Expeditions), Felicis Ventures, Sequoia, Menlo Ventures, General Catalyst, CRV, Amazon, SV Angel, and Carnegie Mellon University.
Investor Quotes
Stephanie Zhan, Partner at Sequoia: “Skild is the emerging market leader, defining the frontier of general-purpose intelligence capabilities in robotics. Deepak and Abhinav’s life’s work is to bring superintelligence to our physical world, and deliver real impact with customers in real-world applications. The trajectory of our progress is once-in-a-lifetime.”
Raviraj Jain, Partner at Lightspeed: "Skild’s foundation models are truly generalizable across form factors, already showing emergent capabilities and are extremely robust - they represent a new paradigm in embodied AI. Unlike several other robotics demos that are often overfitted for the specific demo environment, Skild robots truly work 'in-the-wild,' safely navigating and co-existing with humans."
Sri Viswanath, Head of Coatue Ventures: “Skild AI has built a brain for every robot, trained on vast simulated and real-world data. I believe its rapid progress shows robots are about to transform every industry and sets a bar the rest of the field is still chasing.”
About Skild AI
Founded in 2023 by two pioneers in the field of self-supervised and adaptive robotics, Skild AI is building a scalable foundation model for robotics that serves as a shared, general-purpose brain across diverse robot embodiments and a variety of real-world applications. The company has offices in Pittsburgh and the San Francisco Bay Area. For more information, visit https://www.skild.ai./
View source version on businesswire.com: https://www.businesswire.com/news/home/20250729431330/en/