AI Krishi Sahayak

Inspiration

Indian farmers struggle with fragmented information, unpredictable weather, and complicated subsidy processes. These issues lead to poor crop planning, low productivity, and missed government benefits. We wanted to build an AI assistant that does more than answer questions — it executes end-to-end workflows in farmers’ own languages.

What it does AI Krishi Sahayak is a virtual farming assistant that: • Understands farmer queries in Hindi/local languages (voice/text). • Fetches localized weather and market data. • Provides personalized crop recommendations with clear reasoning (rainfall, soil, MSP). • Auto-fills government forms using profile, Aadhaar, and land records (via OCR). • Generates ready-to-submit PDFs and shares via WhatsApp/SMS. • Ensures privacy and explainability so farmers see why a recommendation or scheme was chosen.

How we built it • IBM Granite models + Agent Development Kit (ADK) for multilingual understanding and agentic execution. • Integrated weather APIs, market price feeds, and subsidy databases. • Designed a mobile-friendly + voice-enabled UI for accessibility. • Added OCR-based auto-fill modules for government forms. • Implemented data security with encryption, consent, and auto-deletion.

Challenges we ran into • Handling multiple dialects and regional language variations. • Designing trustworthy, explainable AI outputs. • Integrating with fragmented government subsidy workflows. • Making the system work in low-connectivity rural areas.

Accomplishments that we’re proud of • Reduced subsidy application time by 90%. • Enabled farmers to access schemes and crop insights in their own language and voice. • Built a scalable, modular framework adaptable across states. • Turned Generative AI from Q&A into real-world execution.

What we learned • Farmers prefer simple, explainable recommendations over black-box outputs. • Workflow automation is the real driver of AI adoption in agriculture. • Privacy and trust are critical in governance-related solutions.

What’s next for AI Krishi Sahayak • Expand to more Indian dialects and languages. • Add AI-powered pest/disease detection from images. • Integrate with crop insurance and credit services. • Create farmer-to-market linkage modules. • Extend the model as a blueprint for digital agriculture globally.

Built With

  • aadhaar-ekyc
  • aes/rsa-encryption-(data-security)
  • agmarknet/msp-feeds-(market-trends)
  • docker/kubernetes
  • government-apis-(digilocker
  • hugging-face-transformers
  • ibm-watson-speech-to-text-&-text-to-speech
  • javascript-(react/react-native)
  • languages:-python
  • lime/shap
  • mongodb-(ocr/unstructured-data)
  • node.js-?-frameworks-&-ai/ml:-ibm-granite-models-+-agent-development-kit-(adk)
  • opencv
  • redis-(caching)-?-apis:-imd/openweather-(weather)
  • sms-gateway-?-databases:-postgresql-(farmer-profiles-&-schemes)
  • subsidy-portals)
  • tensorflow/pytorch
  • tesseract-ocr-?-platforms-&-cloud-services:-ibm-cloud
  • twilio/whatsapp-business-api
  • whatsapp/sms-apis-?-other-technologies:-oauth2.0/jwt-(authentication)
Share this project:

Updates

Submission history