Project Kisan Bheema

Smart farming with AI and Agents: real-time Crop Diagnosis and monitoring, scheme navigation, and market insights for farmers using IoT and intelligent Ai and agents

Comment

Inspiration

In India, especially in rural areas, farmers often rely on outdated methods and intuition to make critical decisions about irrigation, fertilization, pest control, and market sales. With limited digital access, poor internet, and lack of real-time expert advice, this results in reduced yields, resource wastage, and income loss. We were inspired to bridge this gap using AI, IoT, and voice-based technology that can work in local languages, offline, and still provide personalized, actionable guidance — just like a real farming expert.

What it does

Project Kisan Bheema is an AI Agent-powered, IoT-enabled agricultural assistant that helps farmers: 📸 Detect crop diseases from images using Gemini AI 🎤 Ask farming questions via voice in Kannada, Telugu, or Hindi 📊 View real-time sensor data (soil, temp, humidity, NPK, etc.) 📈 Get mandi (market) price predictions 🏛️ Navigate relevant government schemes based on needs 🔔 Receive smart alerts for irrigation, disease, and subsidies 📶 Use the app even in low-connectivity or offline zones

It brings together AI, voice, sensors, and a multilingual dashboard into one seamless smart agriculture platform.

How we built it

We developed Project Kisan Bheema in modular layers: IoT Layer Sensors: DHT11, FC28 (soil moisture), NPK, pH, EC, BH1750 (light) Gateway: ESP8266 + MQTT for real-time communication AI/ML Layer: Gemini for image & text understanding Vertex AI Speech-to-Text and Text-to-Speech for voice input/output ML models: MobileNetV2 (disease detection), RandomForest (crop recommendation), Regression (market prediction) Backend & Data: Python for preprocessing sensor data Firebase Firestore for real-time and offline data storage Firebase Cloud Functions for backend logic and automation Frontend: React Native for the mobile app React.js for web dashboard Hosted on Firebase Visuals: Chart.js, D3.js

Challenges we ran into:

Building multilingual voice interfaces that accurately understood rural dialects Ensuring disease detection from noisy or blurry plant images Working with limited local datasets and training custom models Debugging real-time sensor drift and connectivity issues in field conditions Designing a UI simple enough for non-tech-savvy farmers

Accomplishments that we're proud of

✅ Built a 90% working prototype that runs offline with real sensor data ✅ Integrated Gemini + Vertex AI into a real-world farming scenario ✅ Created a multilingual voice assistant tailored for rural India ✅ Achieved end-to-end flow: photo → diagnosis → recommendation → alert ✅ Successfully tested in simulated rural conditions with low internet

What we learned

How to build Agentic AI workflows using Vertex AI and Gemini Practical edge computing and sensor calibration for agriculture Designing user-first UIs for non-literate audiences Handling real-world agricultural datasets and adapting ML model The importance of empathy-driven design in solving rural challenges

What's next for Project Kisan Bheema

🤝 it should be work in offline 🛰️ Adding drone and satellite-based crop health analysis 🌐 Blockchain integration for farm produce traceability 🧠 Federated learning to improve localized AI performance 📡 Scaling to smart villages, kiosks, and agriculture cooperatives 📲 Adding WhatsApp bot and smart IVR for voice-only farmers

Built With

  • ai
  • bh1750
  • cloud-functions
  • co2
  • d3.js-(data-visualization)-pandas
  • ec
  • fc28
  • firebase
  • firestore
  • gemini
  • github
  • google-cloud-platform-vertex-ai-firebase-(firestore
  • government
  • hosting
  • html/css
  • javascript
  • keras
  • keras-(ai-models)-??-cloud-platforms-&-services-google-cloud-platform-vertex-ai-firebase-(firestore
  • languages-python-javascript-typescript-html/css-??-frameworks-&-libraries-react.js-(web-dashboard)-react-native-(mobile-app)-flask-(data-processing-api)-express.js-(sensor-integration-backend)-chart.js
  • messaging)
  • messaging)-gemini-pro-(multimodal-model-for-vision-+-language)-firebase-studio-?-iot-&-hardware-esp8266-(iot-gateway)-sensors:-dht11
  • npk
  • numpy
  • openweathermap
  • pandas
  • ph
  • python
  • react.js-(web-dashboard)-react-native-(mobile-app)-flask-(data-processing-api)-express.js-(sensor-integration-backend)-chart.js
  • scikit-learn-(ml-&-preprocessing)-tensorflow
  • twilio
  • typescript
Share this project:

Updates