The Simulation Problem Nobody Was Selling
There is a moment in every autonomous systems project when the engineers realize the real problem is not the algorithm - it is the data. Not too little data. The wrong data. Sparse, expensive, dangerous-to-collect, impossible-to-repeat data. A drone that has to fly into a blizzard to understand a blizzard. A self-driving truck that has to be in five places at once to learn five edge cases. A humanoid robot that needs a million falls before it learns to not fall. You cannot fix this with more cameras. You fix it with simulation.
Ian Glow understood this viscerally before most of the world had started using the word "autonomy" in polite company. At Tesla, he was deep inside one of the most consequential simulation programs in the history of engineering - the Autopilot simulation team, where virtual miles drove the real learning that made Tesla's self-driving stack credible. He was not an observer. He was building the infrastructure.
What he saw at Tesla was that simulation done right is not a shortcut. It is a force multiplier. The right virtual sensor gives you data fidelity that real-world collection cannot match. A photorealistic LiDAR model in a procedurally generated city can test edge cases that happen once every hundred thousand miles - and you can run them a thousand times before lunch.
He also saw something else: nobody was selling this capability to the rest of the industry. Tesla built theirs in-house. Cruise built theirs in-house. Nvidia had pieces of it. But there was no platform that let a drone startup in Seattle, or an agricultural robotics team in Iowa, or a green energy firm testing autonomous inspection equipment - there was nothing that gave those teams the same simulation infrastructure that Tesla had quietly assembled over years of internal engineering.
"This is not our first simulated rodeo."
That gap is the founding thesis of Zeromatter. Not "let's build another simulation tool." The thesis is: build the platform, build it properly, and make it accessible to every team doing serious autonomous systems work - regardless of whether they have Tesla's headcount or Tesla's budget.
What Zeromatter Actually Builds
The Zeromatter pitch is four words: "one platform to build, test, and train anything." Behind those four words is a technical architecture that most teams spend years constructing themselves before realizing they should have just bought it.
The platform runs on four core capabilities. First: sensor simulation. Cameras, LiDARs, radars, ultrasonics - all modeled at a fidelity that produces virtual sensor data close enough to reality to train on. Not "good enough for demos" fidelity. Production fidelity, the kind that closes the sim-to-real gap that has historically made simulation outputs unreliable for training real-world systems.
Second: automatic environment generation. Zeromatter uses Houdini - the same procedural CG toolset behind Avatar and Game of Thrones visual effects - to generate diverse, realistic virtual worlds algorithmically. City streets. Rural highways. Warehouse floors. Agricultural fields. The kind of environmental variety that used to require a team of 3D artists and months of hand-modeling can now be spun up in a configuration file.
Third: multi-agent co-simulation. The real world does not contain single actors. Roads have cars, pedestrians, cyclists, and ambient chaos. Warehouses have multiple robots running concurrent tasks. Airspace has multiple drones sharing corridors. Zeromatter's framework lets multiple agents - vehicles, robots, drones - operate simultaneously in shared virtual environments, which is the only way to test the interaction dynamics that actually matter.
Fourth: execution infrastructure and tooling. The unglamorous part that most teams underestimate. Running simulations at scale means kicking off thousands of parallel jobs, collecting results, diagnosing failures, and feeding outputs back into training pipelines. Zeromatter provides the APIs, SDKs, dashboards, and infrastructure to make that loop fast and programmable rather than an ad-hoc engineering project.
Six Industries, One Platform
Zeromatter does not specialize in one vertical - and that is a deliberate architectural bet. The simulation fundamentals that make a drone trustworthy are the same fundamentals that make a self-driving truck trustworthy. Ian Glow's team built the platform to be extensible across domains.
The Team Behind the Platform
The Zeromatter hiring bar is implicit in the roster: engineers who shipped simulation at Tesla, Cruise, Nvidia, Google, Microsoft, Activision, and id Software. The last two are not typos. Video game studios - particularly id Software, the studio behind Doom and Quake - have been building real-time rendering and physics simulation for decades longer than the autonomous systems industry. When your platform needs production-grade rendering, you recruit from the people who solved that problem at interactive frame rates for mass-market games.
The company website says it plainly: "a once-in-a-lifetime team of experts and rockstars" who "move fast and can do what others can't." That kind of copy usually reads like marketing. At Zeromatter, with a team assembled from that particular set of alumni institutions, it reads more like a statement of technical prerequisite.
By early 2026, Zeromatter has grown to approximately 85 people, based in Mountain View in the shadow of Google's campus - a deliberate geography for recruiting from Silicon Valley's deep autonomous systems and deep learning talent pools.
Technology Stack
The technology choices reveal priorities. Rust for performance-critical simulation infrastructure - memory safety without the garbage collector pauses that would wreck real-time fidelity. Bazel for hermetic, reproducible builds across a polyglot codebase. Argo for managing parallelized simulation job orchestration at scale. Three cloud providers (AWS, GCP, Azure) rather than one - intentional flexibility for customers who have existing commitments to specific providers.
The Career That Built the Founder
Ian Glow's GitHub username - NeverNotGlow - dates back to his University of Southern California days, where his repositories tell the story of a computer science student working through data structures, networking, and game mechanics in Java and Assembly. The username is a playful joke that aged well: he has, in fact, never not been building something.
After USC, he moved into software engineering at Microsoft before finding his way into the autonomous systems world through Tesla. The Autopilot simulation role - Manager of Autopilot Simulation from September 2020 to April 2022 - gave him direct experience running one of the most technically demanding simulation pipelines in existence. Tesla's approach to simulation was unusual: instead of treating virtual environments as a testing shortcut, they treated them as the primary source of ground truth for training and validation. Real-world miles were supplementary. Simulation was the lab.
Career Path
Investors Who Bet on the Infrastructure Layer
The investor mix at Zeromatter is itself a signal. Generation Investment Management - Al Gore's sustainability-focused fund - anchors the green energy use case as a legitimate, not incidental, market. AE Industrial Partners and their AE Ventures arm represent aerospace and defense capital, which is not known for speculative bets. Brighton Park Capital and Linse Capital round out the institutional support. That is not a homogeneous investor base: it is a collection of funds that each saw a different wedge of the same opportunity.
The Domain Shorthand That Says Everything
Ian Glow's email is ian@0m.dev. The domain is 0m.dev - zero matter, rendered in numeric shorthand. His GitHub is NeverNotGlow. His personal site is imglow.com. There is a pattern here: a person comfortable enough with their own brand to encode it into infrastructure and let it run quietly in the background. The same instinct that builds a simulation platform instead of pitching one.
Zeromatter's own line - "this is not our first simulated rodeo" - runs in the same register. It is confident without being loud. It says: we have done this before, at scale, in production, where it mattered. We are not learning on your timeline. Come build with us.
Performance, fidelity, and flexibility in a category all of its own. That is the positioning Zeromatter has staked out - and it is backed by the engineering pedigree to defend it.
Fun Facts & Quirks
- His GitHub username NeverNotGlow dates back to USC - it has outlasted every version of his career.
- The email domain 0m.dev is just the company name, compressed into six characters. Efficient.
- Zeromatter uses Houdini - the VFX software from Avatar and Game of Thrones - to build virtual worlds for robots.
- The platform is deliberately multi-cloud: AWS, Google Cloud, and Azure, so customers are never locked into Zeromatter's vendor preferences.
- Zeromatter's team includes game studio veterans from Activision and id Software - because great rendering is great rendering, regardless of whether the end user is a gamer or a ground robot.
- Zeromatter Technologies was incorporated November 2, 2022 - CIN 7115370, Delaware.
Links & Resources
- zeromatter.com - Official Platform
- Zeromatter Team & Company
- Ian Glow on LinkedIn
- @IanGlow on X (Twitter)
- NeverNotGlow on GitHub
- imglow.com - Personal Site
- Startups to Join: ZeroMatter - Sourced Newsletter
- AE Industrial Partners - Zeromatter Investment
- Ian Glow on Crunchbase
- Zeromatter on Crunchbase