Kisan Sathi

The problem

There is no shortage of agricultural information in India. There is a severe shortage of agricultural information that applies to this field, this soil, this district's current schemes, and the specific thing eating this plant.

Generic advice fails in a particular way here: it isn't wrong, it's unusable. A farmer reading a national fertiliser guideline still doesn't know what to do on Tuesday. The gap between correct information and actionable information is where the whole problem lives.

What I built

A mobile-first companion app that resolves everything against the user's location and land:

• Soil assessment and crop recommendation — what this ground will support, not what grows well generally
• Disease and pest scanning — point a camera at the affected plant rather than trying to describe it in a search box
• Scheme information — which government programmes actually apply here
• Fertiliser calculation — quantities for this plot, not per-hectare tables
• Farm diary and alerts — a record that makes next season's decisions better than this season's

The camera-based scanning is the piece that only became possible recently, and it's the one that changes who can use the app. Describing a plant disease in text requires vocabulary a lot of users don't have in the interface language. Photographing it doesn't.

Where it stands

A working prototype, live and usable. The honest status is that the recommendation logic needs validation against real agronomic outcomes before anyone should make an expensive decision on it — which is exactly the kind of claim I'd want checked if someone else were making it to me.

Why it's here

It's the clearest example of building for a user who is not like me, in a context where being confidently wrong has a cost measured in a season's income. Most of the design work was deciding what the app should refuse to answer.