End-to-end IoT + Cloud + Web + ML system
Real-time hyperlocal environmental monitoring with analytics & prediction
Built for real deployment, not just an academic demo
🌐 Live Dashboard: https://iot-hyperlocal-weather-aqi-dashboard.vercel.app
This project implements a complete production-style IoT pipeline — from sensor-level data acquisition to a publicly accessible analytics dashboard with predictive insights.
Environmental data such as temperature, humidity, and air quality (AQI) is collected using an ESP32-based embedded system, transmitted over Wi-Fi, stored in the cloud, processed by a backend ML service, and visualized using a modern web UI.
- Real-time hyperlocal environmental monitoring
- Cloud-based data storage and retrieval
- REST API backend with ML inference
- Interactive dashboards with charts & maps
- Future-value prediction for temperature, humidity, and AQI
- Independent deployment of frontend and backend
- Temperature (°C) — DHT22
- Humidity (%) — DHT22
- Air Quality Index (AQI) — MQ135
- Calibration applied
- Noise filtering & signal smoothing
- Live device location using Leaflet + OpenStreetMap
- Browser-based geolocation fallback
- Latitude & longitude displayed on the dashboard
- Live value cards for quick monitoring
- Historical trend charts (Chart.js)
- Prediction charts for next-step forecasting
- Stable layout with fixed-height graphs
The dashboard provides a unified view of real-time sensor readings, historical trends, and machine-learning-based AQI prediction through interactive charts and maps.
The following dataset represents the cleaned and aggregated historical sensor data used to train the AQI prediction model. Data includes timestamped averages of temperature, humidity, and AQI values with derived AQI categories.
Prediction logic uses a combination of:
- Trend-based forecasting for temperature & humidity
- LSTM-based ML model for AQI prediction
- Bounded projections to maintain realistic values
Note: The AQI LSTM model is trained on locally collected historical sensor data and is intended for short-horizon trend estimation, not regulatory-grade AQI forecasting.
ESP32 (DHT22 + MQ135)
↓ Wi-Fi
Cloud Data Ingestion
↓
Backend API (FastAPI + ML)
↓
Render (Backend Deployment)
↓
Frontend (HTML/CSS/JS)
↓
Vercel (Frontend Deployment)
↓
End User (Browser)
| Layer | Technologies |
|---|---|
| Embedded | ESP32 DevKit V1, DHT22, MQ135 |
| Connectivity | Wi-Fi |
| Backend | Python, FastAPI, TensorFlow, scikit-learn |
| ML Models | LSTM (AQI), Scaler-based preprocessing |
| Frontend | HTML, CSS, JavaScript |
| Visualization | Chart.js, Leaflet.js |
| Backend Hosting | Render |
| Frontend Hosting | Vercel |
iot-hyperlocal-weather-aqi-dashboard/ ├── backend/ │ ├── model/ │ ├── app.py │ ├── requirements.txt │ └── README.md │ ├── frontend/ │ ├── index.html │ ├── style.css │ ├── script.js │ └── README.md │ ├── docs/ │ ├── report/ │ ├── research_paper/ │ ├── ESP32_Hyperlocal_Weather_ML.pptx │ └── abstract/ │ ├── hardware/ ├── .gitignore └── README.md
- Clear separation of device, backend, and UI layers
- Independent deployment pipelines
- Scalable, cloud-ready architecture
- Maintainable and extensible design
- Optimized for real-world deployment scenarios
Ayush Padmawar
Software, ML & Deployment
LinkedIn
Gandharv Birkurwar
Hardware & Embedded Systems
LinkedIn
Designed, implemented, and deployed as a real-world system — not just a prototype.