AI / Engineering2025
Edge AI Smart Irrigation
Predictive irrigation and plant disease detection, from soil sensors to ML.
Predictive irrigation model trained and validated on physics-based synthetic data
Plant disease classification running on-device via TensorFlow Lite
End-to-end pipeline: Arduino sensors, Raspberry Pi logger, FastAPI backend, Next.js dashboard
The problem
Traditional irrigation wastes water and reacts to problems after they happen. The project set out to build a system that predicts what the plants need before they show stress.
The approach
Split into three layers: sensors on the ground (Arduino + BME280), intelligence in the middle (Random Forest irrigation model + MobileNetV2 disease classifier), and a live dashboard on top. Every layer communicates over a clean serial + HTTP pipeline.
The build
- Arduino sensor sketches (soil, humidity, temperature) with a wiring guide
- Raspberry Pi data logger streaming to SQLite
- FastAPI backend with model inference wrappers (no training deps at runtime)
- Next.js dashboard for live monitoring
- Physics-based synthetic data generator to train without waiting for seasons
Want something like this?
This is the kind of work I do every day. Tell me what you want to build and I'll tell you how we get there.