-
"KrishiMitra — Powered by Team Lazer ⚡ Turning weather, price, and crop insights into timely farm actions for India’s smallholder heroes
Here’s a complete Project Story for KrishiMitra – Farmer Decision Agent in the format from your “About the project” guide:
🌱 Inspiration
In India, 86% of farmers are smallholders with less than 2 hectares. A single mistimed irrigation or poor market decision can wipe out profits for an entire season. Seeing this challenge — and the gap between raw data (weather, mandi prices, crop guides) and farmers’ ability to act on it — inspired us to create KrishiMitra, an AI companion that turns uncertainty into actionable plans.
🚜 What it does
KrishiMitra listens to farmers via voice or text in Hindi, Bagheli, or English, pulls hyperlocal weather forecasts, real-time mandi prices, and stage‑specific crop practices, then delivers clear, when–what–how steps with rationales. It also sends SMS/WhatsApp reminders and adapts advice from farmer feedback.
🛠 How we built it
Using IBM’s Granite LLM for reasoning, the Agent Development Kit for multi-agent orchestration, and APIs for weather and prices. RAG pipelines supply crop‑stage best practices, while preprocessing normalizes units, handles transliteration, and ensures dialect‑friendly output.
⚠️ Challenges we ran into
- Data gaps from rural weather/mandi feeds.
- Preserving clarity across dialects while keeping advice precise.
- Integrating multiple tools into a seamless end-to-end workflow.
🏆 Accomplishments that we're proud of
- End‑to‑end demo from voice intake to actionable checklist.
- Linguistic, cultural localisation for grassroots adoption.
- Transparent, rationale-backed recommendations.
📚 What we learned
- Farmers trust AI more when advice is explainable and in their native dialect.
- Automation works best when coupled with simple, familiar delivery channels.
- Guardrails (like avoiding pesticide prescriptions) build safety and trust.
🔮 What's next for KrishiMitra – Farmer decision agent
- Expand to more crops, states, and languages.
- Integrate satellite imagery for early pest/disease detection.
- Partner with FPOs, agri-startups, and government schemes for scale.
- Add offline-first capabilities for low-connectivity regions.
If you want, I can now condense this into a visually‑layered, judge‑friendly slide so it works as both a spoken story and a deck section without losing impact.
Built With
- agmarknet-api-for-mandi-prices
- express.js-(backend)-ai-&-ml:-ibm-granite-llm
- figma
- for
- github-for-version-control
- google-maps-api-for-geolocation-and-directions
- ibm-agent-development-kit-(adk)-data-processing:-pandas
- indic-nlp-libraries-cloud-&-hosting:-ibm-cloud-(functions
- javascript-(node.js-for-apis)-frameworks:-react-(web-ui)
- languages:-python
- numpy
- object-storage)-apis-&-integrations:-openweathermap-api
- ocr-with-tesseract
- twilio/whatsapp-business-api-for-messaging-databases:-cloudant-(nosql)-for-farmer-profiles-and-advisory-history-other-tools:-rag-pipeline-with-vector-db-(pinecone/faiss)
- ui/ux
