A tomato has no respect for a demo. It can face away from the camera, hide behind a leaf, ripen out of sequence or cling to a stem that must keep producing for months. Ask a robot to harvest one and the fruit becomes a small, wet examination in perception, touch, patience and economics. Brandon Contino chose this examination for his company. He did not choose it first.
Contino arrived at the University of Pittsburgh from Denver with neural engineering in mind. He was interested in neural prosthetics, particularly systems that translate brain signals into the movement of robotic limbs, and imagined that a doctorate might follow. His undergraduate years supplied other paths. He wrote grants to build low-cost, internet-connected water-quality sensors. That work introduced him to water scarcity, then to the blunt fact that agriculture consumes a large share of freshwater resources.
The subject changed from how a brain might control an arm to how a food system might use less water. The robot, as it happened, would return later.
The useful wrong idea
At Pitt, Contino met mechanical engineer Dan Chi. They started a hydroponics club, and Chi had grown leafy greens for local food banks. The pair considered building a vertical farm in Pittsburgh. It had the visual ingredients of an appealing future: stacked plants, local production, controlled conditions and careful water use. Then they visited commercial operations and studied the economics. They could not convince themselves that vertical farming would soon deliver commodity-level prices at meaningful scale.
Walking away gave them a better question. Dutch-style glass greenhouses offered many controlled-environment benefits while using sunlight instead of an industrial quantity of lamps. The business model had decades of proof behind it. Contino and Chi began calling greenhouse owners and operators. Some receptionists hung up. Others connected them. The founders asked what kept these businesses from expanding.
The constraint was especially sharp at harvest. Picking is repetitive, physical work performed on foot in heat. The founders had a plausible division of labor for attacking it: Contino knew computer vision, software and motion planning; Chi knew mechanical design and grippers. In 2017 they began developing what became Four Growers. The name refers both to a fourth agricultural revolution and to produce grown across four seasons.
Move into the greenhouse
The early work was not tidy. Contino has recalled being in greenhouses at 1 a.m., then working afternoons when the interior climbed above 120 degrees Fahrenheit. As Four Growers worked with Canadian operators, members of the team stayed in Airbnbs for months so they could remain near the machines and the people using them. Proximity taught them what a lab could not.
A robot must distinguish ripe from unripe fruit, but recognizing the tomato is merely the friendly part. It also needs to perceive leaves, support wires, stems and neighboring clusters. A bad movement can damage a plant and cost a grower many remaining weeks of yield. The motion planner must find a collision-free path in moments, using hardware economical enough to earn a place in a commercial greenhouse.
Four Growers built its own perception architecture and datasets, then paired them with custom motion planning. For the hand, it chose a vacuum-based soft pick. Pulling the fruit with suction reduces the bruising risk and moves faster than a claw closing around each tomato. Eight stereo cameras on the current platform help provide the view. The result is the GR-100, an autonomous system designed to travel a full row, collect produce and carry 24 crates without requiring a person to supervise each pick.
The packing cart test
There is a difference between an invention that behaves nicely on video and a product that improves Tuesday. Contino’s team learned that harvesting, although technically difficult, did not complete the customer’s job. Someone still had to manage what came off the vine. The company developed a packing cart and patented produce-handling technology so the robot could sort harvested fruit into the grower’s existing flow with fewer interventions.
This is the unglamorous center of the story. Four Growers had to solve cameras, models, calibration and mechanical picking, then notice the crates. The grower evaluates the whole shift. A tomato extracted beautifully and left in the wrong place is a clever nuisance.
By 2023, the second-generation GR-100 was operating with customers in the Netherlands and Canada, including at Syngenta’s TomatoVision facility. Contino said the machines were running daily and that autonomously picked tomatoes had entered retail supply. The company reported a typical harvest rate around 43 kilograms of cherry or grape tomatoes an hour, with a peak around 63 kilograms. The economic point was not that the arm could pick. It was that it could keep pace.
The pace of a commercial pick
Teaching the brain another crop
Tomatoes were a starting curriculum. Four Growers designed the system as a platform, with software that could learn other crops and a gripper that could be swapped for a new physical task. Cucumbers followed. Peppers and strawberries entered the company’s public plans, while lettuce appeared in early concepts. Contino describes the reusable part as an “AI harvesting brain”: teach the machine what the new crop looks like, adapt the end effector and retain much of the underlying perception and planning work.
The robot also sees more than a picker ordinarily records. Repeated scans can become plant-level information, yield heat maps and forecasts. Harvesting creates the data; analytics can return it to the operator as a picture of what is ripening, where and when. In this sense the arm has two outputs: produce in a crate and a growing account of the greenhouse.
There are honest boundaries. Dense, bushy outdoor crops remain difficult for machine vision and movement. Trellised crops offer clearer structure. Four Growers has explored outdoor harvesting after participating in John Deere’s Startup Collaborator program, and Contino has described the field version as a transfer of the existing perception, planning and gripping systems onto equipment pulled by a tractor. The greenhouse rails disappear; the hard-won understanding of the crop remains.
A Pittsburgh operating system
Contino and Chi kept the company in Pittsburgh. The decision connects two practical resources: concentrated robotics talent and reasonable access to commercial greenhouse regions in the northeastern United States and Canada. Their own university record had already tested the ecosystem. In 2018, Four Growers won first place and $25,000 at Pitt’s Randall Family Big Idea Competition, then entered Y Combinator’s Summer 2018 batch.
Contino had practiced institution-building before the startup. His public university record includes leadership of Pitt’s Robotics and Automation Society, which he helped grow from a handful of members to more than 60, as well as the Engineering Student Council, IEEE student activities and the amateur radio club. Years later he returned some of that experience as a mentor to companies in Pittsburgh’s Robotics Factory accelerator community.
In November 2024, Four Growers announced a $9 million Series A led by Basset Capital, with Ospraie Ag Science, Y Combinator and existing investors participating. The money was assigned to a less romantic phase: build more machines, expand deployment and support customers across North America, Europe and Oceania. By April 2026, Pitt reported that the company employed more than 30 people and was extending its activity in Europe and Australia.
The quarterly hobby, the durable mission
Contino once offered a revealingly modest fun fact: he tends to change hobbies every quarter. At that moment in 2023, it was Stable Diffusion and the pleasure of generating outlandish images that would take ages to make by hand. The rotating curiosity suits a career that moved through prosthetics, sensors, hydroponics, computer vision and crop logistics without pretending the route had been planned.
The mission has changed less. Contino talks about making high-quality fruits and vegetables more affordable by giving them some of the automation advantage long enjoyed by corn, wheat and soy. The claim is not that one machine fixes food. It is that lowering a grower’s cost can travel through the system, making controlled-environment production easier to expand and fresh produce easier to supply.
The work still comes down to a small encounter. Cameras inspect a clutter of green and red. Software chooses a route. A soft gripper reaches between stems and removes what is ready. The tomato lands in the correct crate. Then the robot looks again. For a founder who followed a chain of better questions, repetition is the point: curiosity can wander, but a harvest only counts when it returns tomorrow.