A robot that sprays less
Most small farms spray the whole field on a calendar. CropScout drives the rows, looks at the plants, and treats only what needs treating: less pesticide on the crop, less cost to the farmer, less run-off into the ground.
Help us teach a farm robot to see
CropScout is an autonomous robot for greenhouses and small vegetable farms: precision spraying and crop monitoring in the rows where a tractor cannot fit. The undercarriage and the sprayer are being built in America. What we don't have yet is the person who makes the machine understand what it is looking at.
- YOLO26
- Python
- Some C++
- Part-time, 20 h/week
- Equity
- USA + Uzbekistan
What we're building
Undercarriage and sprayer
Designed and built by the team in America: drive, chassis, tank, nozzles, the physical machine.
Perception and control
Nobody owns this yet: the detection models, the data behind them, and the code that turns a detection into a decision the robot acts on.
The role: AI/ML subsystem lead
You would own the perception side end to end and help program the robot itself. Concretely:
- Training and iterating detection models with YOLO26, mostly in Python.
- Building and cleaning the datasets those models learn from. This is a real part of the job, not an afterthought.
- Getting the models running on the machine, fast and reliably enough to act on. Some C++ is likely here.
- Working with the mechanical team in America so perception and the sprayer actually meet in the middle.
You don't need to have shipped all of this before. You do need to be able to work without someone assigning you tasks, because there is nobody to do that.
The honest version
This is a low-budget startup: not a funded company with a runway, but a small team building a machine because we think it should exist. We would rather lose you at this paragraph than three weeks in.
- October
- Expenses may be covered from a shared stipend. It is shared, and it is small: treat it as expenses, not pay.
- Nov–Dec
- Unpaid. No stipend, no salary. This is the part people underestimate.
- Equity
- We can offer equity. It is a new startup, so that equity is worth nothing today and may be worth nothing ever.
- Later
- If this works, the role becomes full-time and paid, and you would already be part of it.
- Honestly
- It may not work out. Join only if you are willing to spend the time knowing that.
People who stay
What has hurt us most is not a lack of skill. It is people who start keen and drift off after a few weeks, leaving a half-finished subsystem nobody else understands. That costs us more than never having started.
So we are looking for at least 20 hours a week from September through December. If you can't see yourself finishing December, please apply to something else with our blessing. Saying no now is more useful to us than saying yes and leaving in October.