Build on the platform the labs are trying to buy.
The scarce thing in physical AI is not talent or compute. It is real-world data and a fleet to test on. We have both, and we are hiring the people who will decide what they become.
The honest argument.
Every frontier lab has reached the same conclusion: physical AI is bottlenecked on real-world data, and the industry is about to spend billions buying it.
You can keep waiting on purchased data. Or you can work with the largest known action-conditioned outdoor dataset, 160+ robots adding to it daily, and customers running the results in production.
The platform exists. What it becomes next is the work of the team we are assembling now, with founder-grade ownership to match.
What you would work on.
World models for terrain
WayFAST is a working world model for outdoor terrain. Take it from predicting traction to predicting the physical world.
Learning from action-conditioned data
A decade of aligned observation and action, and a fleet to close the loop on anything you train.
Edge-first autonomy
Models that run in real time on low-cost compute in 95°F (35°C) heat, not in a datacenter.
The data engine
160+ robots in the field. Design what they collect next season.
New verticals
Carry the stack from plantations into mining, energy, logistics, and beyond.
Who we are looking for.
Researchers and engineers at the frontier of physical AI: world models, embodied learning, perception, and field robotics. People who would rather ship into a plantation than into a demo video.
We are a small, global team. The work happens where the machines are: Southeast Asia, India, the United States, and wherever the next vertical takes us.
Tell us what you would build.
No portal, no form letter. Write to us about what you would do with a decade of embodied data and a fleet to test it on.