Rooted Robotics

Coming out of the PhD I wanted to work on problems with more immediate utility, and applied to aerospace, quantum-computing, and photonics positions. The interview processes took so long that in the meantime I started working with a friend on his robotics project — designing automation for indoor agriculture. I wrote a grant to the USDA to try to solve the labor-cost problem in controlled-environment agriculture, and won it.

The case for growing indoors is compelling: almost no water consumption, no pesticides, year-round growing. But CEA is extremely labor-intensive, because there is almost no automation designed specifically for the industry. I leveraged that grant to raise private capital, and that became Rooted Robotics. By the time the aerospace and quantum-computing offers arrived I already had $750K in funding and the company had too much momentum to back out, so I turned the offers down to see how far I could go on my own.

The first system we built was an overhead gantry-driven robot that used cables to navigate within a plane and the gantry to move between planes — each plane a shelf of plants growing in hydroponic trays. It was a difficult system to design, because every time the robot interacted with a tray or a shelf it induced swinging into the whole structure. I built an inertial mass dampener to cancel those vibrations: weights driven along a linear rail from inertial-measurement-unit data fed into a PID control loop.

That project is also what introduced me to computer vision and neural networks. The robot needed to locate the plants and the racks before it could plant or harvest, and from there I learned to train networks to detect plant growth and predict harvest dates.

The system worked. It genuinely replaced manual labor — a person on a scissor lift picking plants by hand. But growers were reluctant to buy it, because it required them to change their entire growing process and the infrastructure of their farm. What they asked for instead was automation that integrates into the processes they already have, with no expensive changes: high throughput, robust and simple designs they could maintain without a specialist visiting the farm. They wanted seeding, harvesting, and washing automated.

That is why we changed the product line to the machines on this page. We made the pivot in the fall of 2024, and the market responded immediately — adopting the technology and giving us more business than I could handle alone. To meet the demand I scaled up the engineering team, bringing on interns and hiring full-time mechanical and software engineers.

Since starting the company we have grown to $1.1M in annual realized revenue at an average margin of 60%, with machines across the US, Canada, the Caribbean, Europe, and Asia, and CE certification for the line. We are now negotiating an acquisition with one of the largest players in the industry, and expect to finalize it before 2027.