“In the physical world, mistakes are permanent.”
Volumes, a spatial data company for physical AI, is built on a blunt piece of arithmetic: of everything humanity has ever digitized, under half a percent describes physical space and how it moves through time. Every earlier wave of AI inherited a data corpus that already existed. Language models had the internet. Computer vision had a century of photographs. Physical AI, the robots and autonomous machines now leaving the lab for the real world, inherited almost nothing. Depth cannot be scraped and friction cannot be downloaded. Someone has to go out and record reality, calibrated and on purpose.
This techno-philosophical thesis threads together the company. Volumes captures real environments with various sensors and records them from every viewpoint at once, a technique called volumetric capture, producing what the company calls “ground-truth reality in motion.”
The founders describe their output as superintelligent data, a record of the most authentic human experience, delivered in emerging formats they are pioneering, such as 4D Gaussian Splatting, or 4DGS. Reality is spatiotemporal, three-dimensional space unfolding through time, and that human reality is not represented in the datasets AI learns from today. With enough of their data at scale, they argue, AI finally “touches grass” and learns how spatial relationships, temporal dynamics and causal physics actually work in the real world. This is the raw material for what researchers call world models, AI systems that learn an internal simulation of how reality behaves, and Volumes positions itself as their spatial data layer. “In simple terms, this data is a bit like memories for the robot brain,” says co-founder Tyler Raciti, “not too different from how a human baby learns the world around her.”
They are not stopping at capturing space and time. The founders want to infuse this data with sensory experience, some of it human and some of it beyond human: the five human senses layered in for a complete picture of a person moving through the world, alongside experiments with modalities borrowed from nature, sonar as a bat uses it, infrared as a python senses it. They call these “non-human variables,” and infused into a spatiotemporal dataset they yield what the company considers the most potent training data in existence, a superintelligent record of physical reality. “How can we ever achieve physical AGI,” Raciti asks, “without a grounding in the Truth of the physical world?”
For Raciti and his Volumes co-founder Chet Ellis, the art of creation is almost a reflex. That said, the two arrived at that problem from strikingly different directions.
Raciti grew up on the old-money North Shore of Long Island, “the area of the Great Gatsby,” Raciti says, “only I didn’t live like Gatsby.” Subtly referencing Jacob Riis, he describes living “how the other half lives,” growing up below the poverty line with his single father. His childhood was unstable. He recalls moving repeatedly, often lacking food, and passing through the juvenile justice system. “My upbringing was a unique duality of status, showing me two worlds simultaneously, but with no bridge between them.” His father, however, supplied another inheritance: an entrepreneurial imagination. He was a musician, artist and cartoonist, the kind of parent who would kick off a father-and-son vagabond magic show touring the country, an early hint of a life spent trying to make things exist that did not before.
At 17, Raciti turned that instinct toward public service. He founded PRO-NEO, an overdose-reduction nonprofit focused on the opioid crisis, and organized a coalition that included educators, emergency responders and parents who had lost children to overdoses. He would later propose legislation that changed New York State law, mandating Narcan, an opioid antidote, in all public school districts.
At 18, he ran for elected office, becoming one of the youngest candidates to run in New York State history. He notes with a smirk that the first person he ever voted for was himself. Beneath that ambition was a lesson he carried forward: “My coming of age was learning to enact change.”
Ellis came to creation through a different household. Born in Santa Monica and raised between Los Angeles and New York, he grew up discussing films with his father, the novelist and screenwriter Trey Ellis, a Columbia University professor whose credits include The Tuskegee Airmen. His earliest organization was an elementary-school chess club whose membership consisted entirely of Ellis, leaving him to play the librarians. In high school he founded a Black Student Organization, and his essays on race and privilege twice took first place in his town’s annual essay contest, picked up by the Associated Press and covered nationally. On the track he set a state-record high jump, clearing seven feet, and was named an All-American before jumping for Harvard, where he studied philosophy and sociology and wrote an honors thesis on incongruity and absurdity. Camus, Sartre and Kierkegaard helped sharpen an idea Ellis had already been practicing: a person could choose what kind of life to build.
After Harvard, Ellis worked on multimodal AI data pipelines at Innodata. There he developed a conviction about a problem he saw building across the industry: AI-generated and human-created material being mixed together without distinction. His metaphor for the problem is characteristically dramatic: “It felt like the Library of Alexandria was burning.” So he left to work on it himself.
Then he spun a globe.
It landed on South Korea, and Ellis began traveling through 14 countries, teaching, filming, writing and collecting material from places he believed were poorly represented in AI training corpora. In Tajikistan, strangers sometimes asked whether his afro was a wig or whether he was Michael Jordan. “First of all, he’s bald!” Ellis recalls answering. The humor sits alongside a harder chapter. After returning home, he says he lived out of his car while trying to build a new approach to data, relying at times on friends at major technology companies for meals.
Raciti and Ellis collided at an NVIDIA event. Raciti was working at Deloitte in autonomous solutions; Ellis was trying to turn his data thesis into a company. Raciti knew the problems from inside the industry; Ellis had been building the solution. They partnered, joined soon after by a third co-founder, Ed Wu, a data scientist who had just left his previously acquired startup and brought the deep technical expertise in advanced hardware and pipeline infrastructure the company needed.
Today their argument is that physical AI, sometimes called embodied AI, has a problem language models never had to solve in quite the same way. A chatbot can generate a bad citation and inconvenience a student. A machine acting autonomously in the physical world can make a mistake with physical consequences. Raciti uses the example of an autonomous vehicle misreading a red light or failing to recognize children near a stopped school bus. “That is what keeps me up at night,” he says. The industry’s emerging answer is a pipeline technically called real-to-sim-to-real, or R2S2R: machines rehearse inside simulations built from real recordings before they ever act in the world, narrowing what engineers call the sim-to-real gap. That loop is only as strong as the reality data feeding it.
It is a revealing anxiety for someone who says successful founders need ambition “bordering on delusion.” Raciti and Ellis are trying to build something difficult enough to require that confidence while working in a field where confidence cannot substitute for accuracy. Their story is what happens when the impulse to create meets the responsibility for what creation can do once it leaves the screen and enters the world. That tension runs through the conversation that follows.
Yitzi: Tyler, it’s so amazing to meet you. Before we dive in and talk about your amazing work, our readers would love to learn about your personal origin story. Can you share with us the story of your childhood, how you grew up, and particularly the seeds and the genesis of all the amazing creativity that has come since then?
Tyler: It’s a lovely question. Let me start with my childhood. I was born and raised in Nassau County on Long Island, in the kind of ritzy, old-money town The Great Gatsby was written about. I lived where the other half lives, except I was below the poverty line with my single dad. It was a rough upbringing, but my dad had a wildly entrepreneurial spirit. He was a musician, an artist, a cartoonist, one of the most interesting people I’ve ever known, and that rubbed off on me.
When I was eight, I performed a magic trick for him. Any dad would’ve said, “Great trick, Tyler.” He said something most dads wouldn’t: “Why don’t you go perform that, and why don’t I join you as a father-and-son magic show?” So we traveled the country doing magic. By eight years old, I was a vagabond magician.
Underneath all of that, my childhood was unstable. We moved every year, and food on the table wasn’t a given. I faced adversity in ways most people don’t and shouldn’t have to, including going through the juvenile justice system. But that instability is exactly what built my resilience. It taught me how to take a hit and keep building.
As a teenager, I founded a 501(c)(3) nonprofit called PRO-NEO, the Overdose Reduction Group, to fight the opioid crisis in New York through legislative and activist work. The short version: I helped get legislation passed in New York State mandating naloxone and opioid reversal drugs across the state.
You might ask how a kid who came from nothing gets legislation passed. The honest answer is that I didn’t do it alone. I built a coalition, an ecosystem of school superintendents, EMT chiefs, and mothers who had lost their children to overdoses. And I learned something that has guided me ever since: as I was lifting all of them up, I was lifting myself up too. My own upward mobility came from creating something bigger than myself. That was my first real glimpse of what became one of my deepest core values, the art of creation. Creating is what lifts me up.
All of that work drew media attention, and it carried me into running for elected office at 18, one of the youngest candidates in New York State history. My fun fact: the first person I ever voted for was myself, because I believed that much in what I was doing. I like doing the hard things. That nonprofit went on to save a lot of lives across New York, the legislation expanded to California, and others have replicated the model since.
Fast forward: I put myself through college and went into consulting. Honestly, consulting was my search for the stability and structure I never had as a kid. Once I found it, once I’d worked alongside autonomous solutions executives and built strategies for frontier physical AI, I realized something was missing, and it was the art of creation. Consulting takes something that exists and scales it. Founding takes something from zero to one, not to be too Thiel-esque. That realization pulled me back to creating, and to one of humanity’s hardest frontiers: physical AI and robotics.
Yitzi: Chet, can you please briefly share your origin story? Tell us a bit about your childhood, how you grew up, and the seeds for all the creativity and ingenuity that came after that.
Chet: That’s a very generous framing of the question. I grew up bicoastal — born in Santa Monica, lived in LA for a while, and then New York. My dad is a screenwriting professor at Columbia University with a few Emmys. He loves writing and movies, and I love filmmaking; growing up, we used to dissect movies together every day.
Through middle school, I was always a big nerd who loved starting things. I started the chess club in elementary school where I was the only member, so I would play chess with the librarians. We still called it chess club, and that was my first time creating something.
In high school, I founded the Black Student Organization at my school. I experienced systemic obstacles that are fairly well-documented; I was featured in articles and press coverage after writing essays on privilege, what it means to be Black in America, and how hard I had to work. Everything came together well for me — I became an All-American high jumper, clearing 2.14 meters, and then attended Harvard College.
At Harvard, I studied incongruity and absurdity within philosophy and sociology on an honors track where I wrote a thesis. I loved Albert Camus, Jean-Paul Sartre, and the existentialists, as well as theology. Looking into the past, even religious existentialists like Søren Kierkegaard — Fear and Trembling really impacted me — made it clear that I could choose my own path in the world.
After graduating with a lot of ambition, I worked at an AI company, Innodata, focusing on multimodal AI data pipelines. I developed concerns about how data projects were being handled across the industry. It felt like the Library of Alexandria was burning because AI-generated content was being mixed directly alongside human work without any distinction. I saw that as an existential issue for the future, where the bias toward rushing training data could lead to serious problems.
So I quit my job to collect data myself. I spun a globe, landed on South Korea, and started traveling across 14 countries. It was an incredible experience. I worked as a teacher, inspired by Jack Ma’s thoughts on how teaching shaped his leadership skills. Managing a classroom of second-graders definitely taught me leadership: how to take a room full of boundless energy and point it toward one shared goal. It turns out that’s great practice for building a startup.
I continued traveling and collecting data — written content, films, and documentation of everything I saw — knowing that regions like Tajikistan were entirely underrepresented in training corpora. Almost nobody goes to Shahriston, Tajikistan, and certainly very few people with an afro. People frequently asked if I was wearing a wig or if I was Michael Jordan, to which I’d say, “First of all, he’s bald!”
When I returned, my parents didn’t quite see the vision of what I was taking on, and I ended up living out of my car while pursuing this dream of revolutionizing the data industry. I would meet friends at Google and other major tech labs who loved hearing my ideas, and they fed me meals as I bounced ideas off of them to shape an enduring, scalable institution.
One of the most interesting people I met was Tyler at an NVIDIA event. He understood the vision immediately. He was working at Deloitte at the time, and when I asked if he wanted to be part of this, he immediately agreed. Then I met our other co-founder, Ed, who is a brilliant data scientist with the deep technical expertise we needed. The rest is history.
Yitzi: You probably have some amazing stories from your career across all these different chapters. I’m sure it’s difficult to single them out, but can you share two stories that stand out in your mind from your career?
Tyler: The first is happening in real time. At Volumes, the physical AI data startup we founded, we’re creating superintelligent data, essentially inventing the memories for robot brains. And as we’re talking right now, I’m at a summit where humanoid robots are walking around me. Most people don’t realize this future is already here. Getting to shape it, and to work on one of humanity’s hardest problems, is deeply meaningful to me.
The second is more personal. My dad passed away when I was 19. He never fully understood what I was doing with my nonprofit, or why a teenager was running for office. But he came to a rally where I spoke in front of a thousand people, asking them to help pass legislation that would make a real difference in New York State. Afterwards he walked up to me and said, “I didn’t understand what you were doing before, but now I understand.” He’d seen a room full of people moved. In that moment I learned something I’ve carried ever since: people don’t have to understand what you’re doing right away. Eventually, they’ll see the impact.
Yitzi: There’s a saying that sometimes our mistakes can be our greatest teachers. I love that. Do you have a story about a humorous mistake you made when you were first starting Volumes, and the lesson you took away from it?
Tyler: You don’t know what you don’t know. That’s the honest answer. There’s real humor and humility in admitting how much we had to learn, and at the same time you need a level of confidence and determination to even begin. So I wouldn’t call any of it a mistake in the traditional sense. It’s the price of admission to building a startup. You take all that unrelenting hustle and ambition and channel it into something greater.
It comes back to what I said earlier about the art of creation: the whole point is building something bigger than yourself. To do that, you almost need a healthy degree of ambition bordering on delusion, because if you could see every single obstacle in advance, you might never start. That mindset is how anything new gets off the ground.
Yitzi: Let’s talk about Volumes. From what I understand, you work with AI and robotics. Tyler, tell us what you’re doing differently from other robotics companies — for example, compared to what Tesla is doing with robotics.
Tyler: We are creating superintelligent data, the most potent form of data on the market.
To step back for a second: AI is, at its core, mechanized learning, and that’s an incredible feat. Philosophers going back to the ancient Greeks spent centuries theorizing about what it means to learn. We mechanized it. We built architectures and fed them the internet. Fed them text, and we got large language models like ChatGPT. Fed them video from YouTube and Vimeo, and we got video generation models.
The next phase is moving AI out of the digital world and into the physical one. A robot is an AI model that can perceive and act in physical reality; the model is the robot’s brain. And the problem today is that robot brains have no memories and no deep understanding of the physical world.
We are collecting real-world data to solve exactly that. Data is neutral until it’s put to use, so what matters is collecting the right data, high-potency data that truly represents human experience. Our approach is orthogonal to conventional methods and very hard to execute, which is why so few companies in the world are doing it. Chet, want to expand?
Chet: Our core thesis centers on four-dimensional data: 3D space evolving over time.
For autonomous, physical AI systems, approximating depth is only the beginning. What does it truly mean to know where a finger will be before it arrives? How do you prevent a glass from falling off a table, or reach past a Tupperware container to grab a specific piece of tin foil? Those are unsolved challenges in data.
Waymo has collected driving data for 17 years, but driving operates within relatively rigid constraints — vehicles move forward, backward, or sideways on flat ground without the unpredictable movements of a child. Deploying a robot into a home or around people requires a far deeper understanding of spatial dynamics over time to ensure safety.
What sets Volumes apart from Tesla’s Optimus or Waymo is that we are building data for the next generation of robot builders. Rather than designing the physical robotic hardware ourselves, we are data specialists equipping the best builders with the best tools.
Yitzi: Keeping the law of unintended consequences in mind, what are some potential unintended consequences we should consider with your technology, especially given its potential to change the world?
Tyler: There’s a major difference between the digital world and the physical one, and it comes down to the cost of a mistake. If ChatGPT hallucinates a citation on a school report, a student gets embarrassed. If an AI video generates an extra finger, we laugh. But if an autonomous vehicle hallucinates a green light with a school bus ahead, or fails to see a stop sign while children are crossing, there is no undo button. In the physical world, mistakes are permanent.
The real world is messy, and it can never be fully modeled from a lab. If AI models aren’t trained on true physical reality and real human behavior, we are building serious safety risks into the foundation. That is what keeps me up at night. I’ve been at humanoid robotics summits in Japan and elsewhere, humanoids walking around the floor, and the word “safety” barely came up. That concerns me deeply. This technology deserves a forward-thinking, Star Trek-style vision of the future, one where AI and robotics make society better, responsibly and by design. That’s the future we’re building toward at Volumes.
Chet: Simply put, our goal is to generate information. While information itself does not have an inherent moral charge, there is a moral imperative to expand human knowledge and wisdom.
Expanding that knowledge naturally amplifies the potential for both harm and good. We do not view ourselves as the sole arbiters of that balance, but rather as enablers who empower good people to do good things. We want to support the builders who will use these capabilities for positive impact.
Yitzi: Here is our final aspirational question: Tyler and Chet, each of you has built a significant platform through your work and holds meaningful influence. If you could spread an idea or inspire a movement that would bring the greatest amount of good to the most people, what would that be?
Tyler: I truly believe my life’s purpose is the art of creation. Voltaire said that we must cultivate our own garden, but the broader question is: what are we cultivating that garden for? I view my life through the lens of what I can leave behind that maximally benefits the next generation.
Innovation, especially in physical AI and robotics, has immense potential to help or harm. That is the new frontier. Drawing from my experiences, I believe the most meaningful way to create change is by positively advancing humanity towards a greater order of good and cultivating a garden where everyone, not just a few, can thrive.
Chet: My mind goes to many worthy causes, but the greatest impact comes from inspiring hope and optimism. Every person should recognize that we do this work for the next generation, so that their struggles are fewer than ours. If we can inspire young people by demonstrating that you can achieve tremendous success with integrity, without lying, cheating, or stealing, we will leave the world a much better place.
Yitzi: Tyler and Chet, it’s been an honor to meet you. How can our readers continue to follow your work, engage your services, or support your mission?
Chet: The best place to start is our website, volumes.cloud, where we will share white papers. You can also connect with us on LinkedIn and follow our writing. We love connecting with like-minded people who want to empower others through technology. If that resonates with you, please reach out to us.
Yitzi: Amazing. Tyler and Chet, it’s been fantastic meeting you. I wish you continued success, good health, and blessings, and I’d love to do this again next year.
Tyler: Thank you so much, it was such a pleasure.
Tyler Raciti & Chet Ellis On Volumes, Their Startup Building Superintelligent Data for Physical AI was originally published in Authority Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.