Building edge AI for real farms

Most AI lives in the cloud. My final year project lives in the field. Arduino sensors, a Raspberry Pi logger, and ML models that predict what plants need before they show stress.

The irrigation model is a Random Forest trained on physics-based synthetic data, because waiting for seasons to collect real data isn't an option. The disease classifier runs on-device via TensorFlow Lite, so it works even without connectivity.

Building it taught me the whole pipeline: sensors, serial protocols, model training, model serving, and a dashboard that makes it all visible. FastAPI on the backend, Next.js on the front.

It's the project I'm proudest to show, because it proves the full stack, literally.