Google DeepMind has hired Aaron Saunders, the former chief technology officer of Boston Dynamics, as vice president of hardware engineering, a move that signals a deeper push into robotics. Saunders, credited with work on Boston Dynamics’ back-flipping and performance robots, joined earlier this month.
The appointment supports DeepMind CEO Demis Hassabis’ plan to make Gemini function like an operating system for physical robots, similar to how Google supplies Android to a range of smartphone makers. “You can sort of think of it as a bit like an Android play […] We want to build an AI system, a Gemini base, that can work across any body configuration,” Hassabis said in an interview with WIRED.
Boston Dynamics is best known for legged robots, including four-legged, dog-sized platforms and humanoid systems that have performed remarkable acrobatics. Saunders was involved in an amphibious six-legged prototype before he moved up to VP of engineering in 2018 and then to CTO in 2021.
Google DeepMind has produced notable robotics research for years. Interest in humanoid and other advanced form factors has grown, and researchers at the company are increasingly focused on producing AI models that can command motors, perceive environments, and coordinate complex behaviors on real hardware.
Hassabis said he is enthusiastic about where that work is heading. AI-powered robotics “is going to have its breakthrough moment in the next couple of years, if I was to predict,” he explained, framing the hire of an experienced hardware lead as part of a broader strategy to connect software and machines.
Boston Dynamics is majority owned by the South Korean automaker Hyundai Motor Company. Hyundai bought its stake from SoftBank, which had acquired the firm from Google's parent company, Alphabet, in 2017.
The parts, sensors, actuators, and engineering know-how needed to build legged robots have become more accessible. Several U.S. startups are now pursuing humanoid designs, with names including Agility Robotics, Figure AI, 1x, and Tesla among the most visible. Elon Musk recently said that his company aims to produce a million of its Optimus humanoids over the next decade.
Chinese firms have made rapid progress on cost and volume. Unitree, based in Hangzhou, has pushed price points down and, in recent months, overtaken Boston Dynamics as the largest supplier of four-legged systems for industrial uses such as manufacturing and construction.
Hassabis said he admires Unitree's work but that his prime interest is in software. “I'm most interested in the [AI] brain part of it,” he says, adding that the multimodal capabilities of Google DeepMind’s flagship model Gemini are particularly well suited to robotics.
Saunders brings practical experience at the intersection of mechanical design and control software, expertise DeepMind will need as it moves models from experiments into physical systems. His background with agile legged machines and prototypes gives him familiarity with rapid iteration on motors, sensors, and balance controllers.
Turning a foundation model into a broadly useful robot platform demands work on perception, language grounding, motion planning, and safety. Engineering teams must adapt a single model to different body shapes and sensor packages, then validate behavior in simulation before long-term tests on hardware.
Gemini's multimodal design — combining vision, text understanding, and planning — helps a model take camera input, interpret intent, and propose motor-level commands. DeepMind's researchers aim to shrink the gap between lab demonstrations and machines that can operate reliably in industrial and service settings.