Inspiration

In Indonesia, most farmers still track everything in their heads or in worn notebooks — when they fertilized, sprayed, or harvested. There's no data history, so every decision is guesswork. And when crops get sick, expert help is hard to reach: agricultural officers are few, and diseases destroy harvests before answers arrive.

I asked a simple question: what if every farmer had an agronomist in their pocket? That's why I built BerTani — from the Indonesian word for farming.

What it does

BerTani is a smart farming assistant for Indonesian farmers:

  • Manage land — register multiple plots with photos, crop types, size, and GPS location
  • Log activities by voice — farmers speak naturally ("I fertilized 50 kg today") and AI converts it into structured records; manual logging works too
  • Diagnose plant diseases — snap a photo of a leaf and AI identifies the disease, severity, and treatment steps
  • Field weather & humidity trends — hyperlocal forecasts so farmers know the best time to spray or irrigate
  • Harvest tracking — complete harvest history with total yield
  • Sobat Tani chatbot — an AI agronomist available 24/7 in Indonesian
  • Community & gamification — XP, levels, 6 claimable badges, reputation, and a farmer leaderboard
  • Monetization — premium subscription + coin economy powered by RevenueCat

Everything is designed for real field conditions: dirty hands, one free hand, low literacy barriers — so voice-first input matters.

How we built it

  • React Native + TypeScript for Android & iOS
  • Supabase: auth (with anonymous → Google account migration so guests never lose data), PostgreSQL with Row Level Security, storage buckets, and 7 Edge Functions for all sensitive logic
  • Google Gemini AI powering four features: chatbot, photo diagnosis, voice-to-activity parsing, and content moderation — with a model fallback chain for reliability
  • RevenueCat SDK for subscriptions and coin packs, synced via webhook to keep entitlements consistent
  • Open-Meteo API for hyperlocal weather, with dynamic endpoint configuration
  • Native Android module + WebView for speech recognition
  • Local scheduled notifications (@notifee) for farming reminders based on each user's land data

Challenges we ran into

  1. Turning messy human speech into clean data. Farmers don't speak in forms. I built a Gemini-powered parser that extracts category, quantity, and notes from free-form speech.
  2. AI cost sustainability. Every scan/chat costs money. The solution: coins earned through engagement (daily login, badges) or purchased directly, while premium subscribers get unlimited AI usage.
  3. Reliability. A single model failure can't break the app — hence the multi-model fallback chain across all AI features.
  4. Android lifecycle crashes from fragment restoration that only appeared in production — fixed at the activity level.
  5. Safety first. AI advice about pesticides can affect livelihoods, so moderation and prompt guardrails came before features.

Accomplishments that we're proud of

  • Shipped a fully working app on Google Play during the contest window — not a prototype
  • Voice logging actually works with natural Indonesian speech, not rigid commands
  • A complete monetization stack: subscriptions, consumable coin packs, rewarded ads, and interstitials — all through RevenueCat
  • Gamification that makes daily record-keeping feel rewarding instead of like a chore

What we learned

  • Shipping early beats perfection — releasing during the contest forced fast iteration based on real user feedback
  • AI products need graceful failure modes; users trust an app that degrades gracefully more than one that's "smart" but fragile
  • Designing for your users' physical reality (one hand, sunlight, noise) changes every product decision
  • Monetization must fit the audience: smallholder farmers won't pay upfront, but they will pay for value at the moment of need

What's next for BerTani

  • Automated care schedules predicted from crop type, weather, and activity history
  • Exportable reports (PDF/CSV) so farmers can access subsidies and financing
  • Deeper community features: forums, following, and knowledge sharing between regions
  • iOS release to reach more farmers
  • Partnering with local agricultural agencies to bring verified disease data into the Plant Clinic

Built With

Share this project:

Updates

Submission history