Arima

Harvesting robots for small farms, paid per acre.

We're building a machine that cuts soybean and pulses on small plots in Maharashtra, and charges the farmer the way labour does: by the acre.

What the rover sees in simulation. Green is plant, the red line is the gap it steers toward, the dotted white line is the centre of the camera image. The distance between them is the steering error.

The harvest doesn't wait for labour

In Marathwada, soybean is cut by hand in a window of about ten days before the pods shatter. There are fewer farm workers every year, and the ones left have other options, so at harvest the cutters often arrive a week late, and they set the price. Farmers wait, lose pods, and this year some ploughed the crop under.

Rented reapers remove the hard work but not the dependence: they need a skilled operator, who is scarce in the same window, and they cut at a fixed height, so short and fallen plants are missed.

₹3,500–5,500
per acre to cut soybean by hand, 2026, Latur district
7–10 days
the harvest window before pods shatter
6–8 days late
how long farmers told us they typically wait for cutters

What we're building

A compact field robot sized for small, irregular, intercropped plots. It drives itself along the rows, cuts at the right height for each crop, and runs long hours, including the cool early morning when pods shatter least. One village operator books the jobs and looks after several machines.

The farmer pays about ₹3,000 an acre, less than hand labour, and the machine shows up on the day it's booked. The same base takes a seed drill and an inter-row weeder, so it works more days a year than a harvester alone.

Where we are

Pre-revenue, pre-hardware, and honest about it. The simulation works: the rover finds the rows with its camera, steers down the gap, recovers when it's knocked off line, and stops at the end of the row.

Gazebo simulation, October 2026. The rover model carries a camera, IMU, GPS and LiDAR; row detection is colour-based, which works on green plants and will need a trained model for brown harvest-ready crop.

  1. 30+ farmer interviews across 5 villages in Latur district, with per-acre cost data
  2. Working simulation: row detection, steering, recovery, stop at row end
  3. Physical rover: parts [ordered DATE], assembly in Bangalore
  4. First paid acres with an operated reaper, February 2027 chickpea harvest
  5. Rover on real rows, then autonomy on the reaper

Who's building it

Avinash Shinde, founder. Grew up in Latur; the family farm is 40 km away and his mother farms it today. Nine years of software close to hardware: printer R&D at HP, then four years at Esper building fleet management and real-time remote control for about a million Android and Linux devices, including robots on Jetson. Based in Bangalore, moving to Pune to be four hours from the fields.

Looking for a cofounder

We want a technical cofounder who has built and deployed robots outdoors: mechatronics, plus perception and control. Equal or near-equal stake with standard vesting. You'd be in Bangalore for the build and in the fields of Marathwada for every sowing and harvest window.

hello@arimalabs.xyz