Grow with us
Teaching machines to farm — with a human always in the loop.
Farm robots and drones are trained in laboratories on tidy, flat, single-crop fields. Real farms in India, Nepal and Tibet are terraced, intercropped, flood-hit and unpredictable. So we train the machines where the work actually happens, with volunteers, farmers and engineers standing right there.
How the training works
Four stages, every one supervised
A machine only earns more independence after a human has watched it do the job correctly, season after season, on that specific farm.
1. Ground truth
Volunteers walk the plot with phones and drones, capturing crop rows, weeds, bunds, slopes, water channels and obstacles. Every image is labelled with what it actually is.
2. Shadow runs
A machine attempts the task — weeding, seeding, spraying, silt clearing — while a trained human walks beside it holding the stop switch and correcting every mistake.
3. Edge cases
The odd cases are the whole job: a stone, a muddy rut, an intercropped row, a child's footpath, a seedling that looks like a weed. Humans flag them so the model learns them.
4. Supervised autonomy
Once a machine repeats the task safely across seasons, it takes over the heaviest, most repetitive work — still watched, still logged, still correctable.
Every discipline is needed
No single field can build a farm robot
A weeding robot that a roboticist builds alone will cut seedlings. A machine an agronomist specifies alone will never survive the mud. Every combination of these fields is useful, and most of them are necessary.
Agriculture & agronomy
Defines what 'done correctly' means: crop stage, root-zone depth, weed versus seedling, canopy health, organic-safe practice, harvest readiness.
Robotics, AI & computer vision
Navigation without reliable GPS under canopy, obstacle avoidance, on-device inference, grippers and cutters that do not bruise the crop.
Mechanical engineering
Chassis that survive mud, dust and 45°C heat, sealing against water ingress, and repairs possible with the tools a village workshop already owns.
Electrical & power
Solar charging docks, battery life and thermal safety, low-power sensors, and rural connectivity where mobile signal drops out.
Civil & irrigation
Contour bunds, terrace load limits, soil compaction thresholds, drainage and drip lines — so a machine never damages the land it works on.
Drone & survey work
Aerial mapping, multispectral crop health scans, flood and silt damage surveys, and flight rules compliance for each country.
Where you fit
Ways to take part
Volunteers & students
No robotics background needed. Label data, walk shadow runs, record field notes, assist pilots. Your hours count toward certification like any other farm work.
Get involvedAcademia & research labs
Use real working farms as a testbed for capstone projects, theses and field trials. Supervise students, validate methods, co-author the datasets that come out.
Partner with usSMEs & equipment makers
Test prototypes on smallholder terrain instead of a lab yard, with farmer feedback after every run and volunteers to help operate and document.
Talk to usFarms hosting a trial
Offer a small plot for supervised trials. You keep full control, you can stop any run, and you get the labour help while the trial is on.
List your farmOur rules for machines on a farm
- Machines take the worst jobs, not people's jobs. We target work that already goes undone or injures people — hand-weeding in extreme heat, desilting after floods, carrying loads on steep terraces.
- A human is always within reach of the stop. Every run has a named supervisor standing close, with a working emergency stop. No unattended autonomous runs on a host farm.
- The farmer decides, always. The farm can pause or end a trial at any moment, and nothing enters a field without the farmer's agreement that day.
- Data about a farm belongs with that farm. Field data is collected with consent, and we aim to publish open, locally relevant datasets rather than lock them away.
- Nothing is claimed until it is proven. Trial results are reported honestly, including the runs that failed.
Your time counts
Training hours are field hours
Time spent labelling, piloting, supervising a run or documenting a trial is logged the same way as any other farm work, reviewed by our team, and counts toward your certification.
Log your field timeStart a collaboration
Bring a lab, a team or a prototype
Tell us the discipline you work in, the problem you want to test and the season you need. We will match you with farms and volunteers who fit.
Tell us about your work