← Back to work

Full-Stack

Munda

An AI crop advisory built for Zimbabwean smallholder farmers, turning local growing conditions into plain language guidance a farmer can act on.

Munda screenshot

The problem

Smallholder farmers in Zimbabwe make the decisions that determine their harvest, when to plant, when to harvest, what disease to watch for, largely on their own. The agricultural extension officers who are supposed to advise them are stretched thin, so timing mistakes that cost a whole season are common. The advice exists, it just does not reach the person in the field.

What I built

  • Built an advisory that turns local growing conditions into plain language guidance a farmer can act on, rather than generic tips.
  • Named it Munda, the Shona word for farm, and shaped it around the reality of the user, who often has limited connectivity and is not a data person.
  • Used Claude underneath so a farmer can ask in plain language and get an answer that fits their crop and their situation.

The outcome

Munda puts the kind of guidance a scarce extension officer would give into the farmer's hands directly, which is the difference between a good season and a lost one.

Stack

  • Next.js
  • TypeScript
  • Claude API
  • Supabase