The AI behind Khet Sathi.
Built with a modern, on-device-first AI stack — because farmers deserve technology that works in the field, not just in the lab.
Classifies leaf and crop images to spot early-stage disease patterns invisible to the naked eye.
Custom CNN architectures trained on 200,000+ labelled Indian crop images across 40+ disease classes.
Models run on-device so disease detection works offline and stays fast, even on entry-level phones.
Powers our low-latency inference and data APIs, with autoscaling for peak sowing seasons.
Large language models drive natural-language answers in Odia, Hindi and English inside the voice assistant.
Multiple providers blended into a hyperlocal, village-grid forecast updated hourly.
Privacy by design
Wherever possible we run inference on-device — your photos and voice never leave your phone unless you explicitly upload them for a second opinion. Cloud data is encrypted in transit (TLS 1.3) and at rest (AES-256).
Continuous learning
Every corrected diagnosis you submit becomes training data (fully anonymised) that improves accuracy for every farmer in the network. Community-powered intelligence.