The workforce is vanishing
+25% wages since 2023
Seasonal labor is scarcer and costlier every year. Growers plan a season around whether a crew will even show up.
Autonomous field robotics for specialty crops
Specialty farming runs on skilled hands — and those hands are getting harder to find every season. We're building autonomous robots to take on the hard, repetitive field work, so growers can keep every row tended, every plant cared for, no matter who shows up.
Autonomous robots, built for the row.
The problem
The crops we love most — grapes, berries, orchards, greenhouse vines — can't be mass-mechanized like grain. They need judgment, care, and a human touch, on every plant, all season long. And the people who do that work are getting harder to find every year.
+25% wages since 2023
Seasonal labor is scarcer and costlier every year. Growers plan a season around whether a crew will even show up.
up to 40% of production
Pruning, thinning, weeding, spraying, harvesting — the bulk of what a specialty crop costs to grow is still done by hand.
every plant is different
Trellised, delicate, and irregular — specialty crops defeat blunt machinery. The work needs to see, understand, and adapt.
every row, every season
Repetitive, time-critical, and relentless. Miss the window and the crop suffers — so the pressure never really lets up.
Grapes, berries, orchards, greenhouse vines — the crops that feed the most people and can least afford to lose their crews.
Why it's still unsolved
Grain got automated because a field of wheat looks the same everywhere — a machine can run it blind. Specialty crops are the opposite. Every vine, branch, and fruit is different, and the work needs a decision on each one. That's not a mechanical problem. It's a perception problem.
crop vs. weed · leaf vs. stem
A tractor follows a line. It can't see which plant needs work and which to leave alone — so it treats the whole row the same and wastes most of the effort.
millions of decisions a day
The value is in judgment repeated at scale — the right choice on this vine, then the next, then the next. Only a person could do that. Until now.
dust, heat, sun, and time
A demo in a greenhouse is easy. A machine that works a real field, day after day, through weather and wear, is the actual bar — and where most attempts fail.
Our approach
We're teaching machines to do what a skilled worker does — look at a plant, know what it needs, and act on it — then repeat that, tirelessly, across a whole field. Three capabilities, working as one.
01 · See
A vision model built for plant geometry reads each vine, leaf, and stem in 3D — not a GPS line on a map, but the plant itself, as it actually stands in the row.
02 · Understand
The model makes the same call a trained hand would — what to do on this plant, and what to leave alone — and gets better with every field it sees.
03 · Act
A rugged, electric field robot carries it out — plant by plant, row by row, all season, unattended. Farmer-first hardware built to survive a working field, not a lab.
Where this goes
The hard part isn't the crop — it's teaching a machine to understand plants at all. Once it can, that understanding carries. A model that masters one task on one crop makes the next task, and the next crop, far easier. Every field we work makes the whole system smarter.
We pick one demanding job on one crop and make a machine genuinely good at it — the honest test of whether the intelligence is real.
The same understanding of plant structure moves from vines to orchards to berries — so each new crop starts from a head start, not from scratch.
Every field feeds the model. The more plants it sees, the better it works — a flywheel that compounds with every acre we cover.
The goal, plainly: every plant on Earth, seen, understood, and tended.
Work with us
Growers, partners, and the curious — tell us about your operation and what wears your crew down most. We're picking our first fields now.