SafeWorld

Culture

Why we are ‘building a SafeWorld’

Our mission is to improve robot safety for the next billion machines. Meet the founders building the safety infrastructure for intelligent robots.

By Kyle Wong, Dr. Ding Zhao, and Simo Rachidi

Today, we’re excited to introduce SafeWorld. Our mission is to improve robot safety for the next billion machines. AI is moving from screens into the physical world. More capable robots require more capable ways to test them.

For decades, industrial robots have largely operated in controlled environments, following predefined instructions. AI is changing that. Robotics is making rapid gains in capability, unlocking machines that can take on dangerous work, expand what people can accomplish, and bring new capabilities into factories, warehouses, hospitals, and beyond.

That shift is opening up a much wider range of possibilities. AI-powered robots can adapt to more complex situations, but their behavior is also probabilistic and harder to anticipate. Understanding how they behave around people requires testing far more scenarios, including rare and potentially dangerous ones. We believe one of the next major challenges is turning those capabilities into products that people can safely deploy at scale.

Physical testing remains essential. But it is slow, expensive, and limited in the range of situations it can cover. Real-world testing alone cannot reproduce every unexpected human movement, unusual interaction, or combination of conditions a robot might encounter.

Our conviction is simple: if autonomy is going to scale, safety has to scale with it. The physical world will need an entirely new infrastructure for trust.

That’s what we’re building at SafeWorld. As Waymo co-CEO Dmitri Dolgov recently put it, “In the physical world, trust is everything.” And as robots become more capable and operate more closely around people, earning that trust gets harder.

We’re starting with one of the most immediate problems teams face today: how to test increasingly capable robots around people at scale. SafeWorld provides robot safety testing and simulation software that lets teams explore thousands or millions of variations of situations that would be difficult, expensive, or dangerous to reproduce physically.

We’re working to turn months of manual testing into an automated workflow that lets teams investigate more possibilities, earlier in development. We’re already working with robotics companies and global enterprises to make this process faster and more systematic.

Done well, this doesn’t just help prevent failures. Better safety infrastructure can give teams the confidence to increase robot performance, operate closer to people, enter new environments, and deploy capabilities that otherwise might never leave the lab.

The opportunity is much broader. As robots become more capable, we expect entirely new infrastructure will be needed to evaluate risk, validate changing AI systems, generate safety evidence, monitor deployed systems, and learn from failures in the field.

At the same time, we want to push the frontier of physical AI safety itself. We plan to develop and open-source new research, benchmarks, and evaluation methods for understanding how increasingly intelligent machines behave in the physical world.

There is a lot left to build.

Why us

Building the safety infrastructure for intelligent machines is going to be a difficult, multidisciplinary problem. It spans robotics, AI and safety research, simulation, perception, large-scale software systems, and the practical realities of deploying autonomous systems in the physical world.

Our team has worked across robotics, autonomous vehicles, AI safety, simulation, enterprise AI, and large-scale infrastructure at organizations including Google DeepMind, NVIDIA, Aurora, Amazon, Salesforce Einstein, and Xiaomi, with research roots at Carnegie Mellon, Stanford, Caltech, Tsinghua, and UC Berkeley.

Robot safety won’t be solved by any single model, simulator, standard, or discipline. It will require connecting advances in robotics and AI with rigorous evaluation, reliable infrastructure, and the realities of deploying autonomous systems in the physical world.

That same combination of research, engineering, and company-building experience starts with the three of us.

Portrait of Ding Zhao, SafeWorld co-founder.Portrait of Kyle Wong, SafeWorld co-founder.Portrait of Simo Rachidi, SafeWorld co-founder.

Ding Zhao

Kyle Wong

Simo Rachidi

Ding has almost two decades of experience working on safe autonomous systems. As a Professor and Director of the Safe AI Lab at Carnegie Mellon University, a former visiting scholar at Google DeepMind, and an NSF CAREER Award recipient, he brings deep research experience in understanding and evaluating risk. He has also contributed to the development of international safety standards, including serving as a U.S. representative.

Kyle is a repeat founder who brings experience building companies and products centered around customer trust. He previously founded Pixlee, an AI and user-generated content platform that worked with thousands of brands, including Nike, Airbnb, Sephora, and Delta, helping companies build trust with customers through authentic content. Pixlee was acquired in 2022. Most recently, Kyle served as CEO of StartX, Stanford’s startup accelerator, where he worked closely with professors, students, and founders building new companies across AI and robotics.

Simo brings experience at the intersection of machine learning, cybersecurity, and autonomous systems. He holds a master’s degree focused on autonomous driving and is a repeat founder and former Principal Security & ML Engineer at Salesforce Einstein, where he worked on pioneering systems at the intersection of AI agents and cybersecurity. He also architected data pipelines and machine learning systems operating at petabyte scale. At SafeWorld, he’s applying that experience to building the infrastructure needed to evaluate increasingly autonomous systems safely and at scale.

We’re fortunate to have investors who share that ambition. Alongside Shine Capital and a16z Speedrun, our investors include BoxGroup, Carnegie Mellon University Endowment, Innovation Endeavors, SV Angel, Ovo Fund, Valkyrie, Zelda Ventures, Alpha Square Group, Founders Future, and Brave Capital, as well as founders and executives across the frontier of AI and robotics, including Waymo, NVIDIA, Google DeepMind, OpenAI, Aurora, Amazon Robotics, Dyna Robotics, and Generalist.

The next billion robots will operate in an extraordinary range of environments. Helping teams understand and evaluate safety at that scale is a challenge, and we can’t wait to bring more people into this effort.

If you’re building or deploying autonomous systems, we’d love to work with you. And if you’re a researcher, engineer, or builder who wants to tackle some of the hardest unsolved problems at the intersection of robotics, AI, and safety, we’d love to build with you.

The physical world is about to become much more intelligent. We want to help make sure it becomes safer at the same time.

Join our beta waitlist

Explore open roles

Want to help us build the future of safe robots?

Explore open roles