A driverless 1930s tractor with a brass sensor mast pulls a seed drill while a second driverless tractor works the next field and a biplane carrying a camera pod passes overhead. A farmer watches from the field edge.

Physical AI

Models that drive machines

From the lab bench to the boom.

We integrate our models with the next generation of farm machines, so every pass across the field is more precise and more sustainable than the last.

Edge survey computer

SOLV: Survey Optics at Ludicrous Velocity.

SOLV is an NVIDIA-based edge computing system with up to six camera inputs, built light enough to fly and powerful enough to run modern models on live data.

Run it on our proprietary acceleration system and optimized models for extreme processing throughput, or bring your own models and pipeline. SOLV mounts on drones and other aircraft, ground vehicles, and farm equipment.

Ask about SOLV

SOLV · Edge survey unit

Cameras: Up to 6 inputs
Compute: NVIDIA Orin AGX or Thor AGX
Software: Precision AI acceleration, or your own models
Uplink: Cellular or Starlink
Geolocation: Millimetre-level
Mounts: Aircraft, ground vehicles, farm equipment

Real-time mapping

The map is finished when the wheels touch down.

SOLV stitches a high-resolution orthomosaic while the vehicle is still moving.

Each frame is analyzed as it is captured. The important results are vectorized and streamed back over cellular or Starlink, and every finding geolocates to its high-resolution source image with millimetre-level precision.

Frames stitched: 0 / 96
Points vectorized: 0
Status: On the ground

Illustration of in-flight stitching and vector streaming.

A 1930s tractor seen from behind carries a wide steel spray boom over rows of young corn. Seven brass nozzles with glass lenses and amber lamps hang from the boom.

Sprayer research

A sprayer that knows the species.

We are building hyper-precise ground sprayers that take sense-and-spray to the level of the individual plant.

Variable-ratio foliar nutrition

Soil sampling tells a farmer what each part of a field needs. Our sprayer applies foliar nutrients in variable ratios to match it, plant by plant. Our research target is a 65% reduction in fertilizer.

Species-level pesticide application

Our models identify weeds by species and detect herbicide-resistant strains. The sprayer applies higher-group herbicides to resistant kochia and waterhemp and to nothing else.

Herbicide resistance

Kochia and waterhemp are outrunning the chemistry.

Herbicide resistance is one of the largest threats to food security. We are developing models to detect resistant strains in the field and eliminate them one plant at a time.

Why it matters

An agronomist in overalls and a flat cap kneels in a soybean field and reads a tall waterhemp plant with a brass handheld analyzer, cabled to a vacuum-tube instrument case on a folding table beside a microscope, sample vials and a field notebook. A tagged kochia plant stands a few rows back.

A riveted aluminium spray drone with wooden propellers flies low over a crop and sprays a single weed. A 1930s farm couple watch from a rail fence.

Drone research

The AirSprayer was the world's first see-and-spray drone.

We built a drone that finds weeds in flight and sprays them precisely. It is now being commercialized by a dedicated company, Precision Aerospace.

The perception models, edge compute, and real-time pipelines built for the AirSprayer carry into SOLV and our ground sprayer research.

A small four-legged herding robot of riveted enamelled steel and brass trots behind a flock of sheep as they file through a gate while a farmer leans on the fence with a cup of coffee.

Autopilot research

Autonomy a farmer can trust with the field.

We conduct primary research into aerial and ground-based autopilot systems, with safety as the first requirement.

Build it with us

We work with equipment makers and research partners who want our models on their machines.

Start a conversation