202 meta successful

Stetho: a stethoscope for every pump, delivered through WhatsApp. The problem. Small farms and workshops depend on electric pumps and motors that nobody monitors. When a pump starts to fail, it usually sounds different for weeks, but the owner has no way to tell whether “different” means “call someone now”. Professional vibration analysers cost more than the pump. Phone vibration apps need the phone pressed against a running machine, which we consider unsafe and will never ask anyone to do. Our insight. A phone is a poor vibration instrument in absolute terms: an uncalibrated microphone, a camera too slow for motor speeds, and heavy video compression. But the same phone, held at the same marked spot, recording the same pump, is far more consistent relative to itself. So Stetho doesn’t try to name a fault from one video. It asks the question a phone can answer: is this pump different from its own healthy normal? How it works. The owner rests their phone near the running pump and sends a 10-second video on WhatsApp. Stetho then: Checks recording quality, and asks for a new video, with specific advice, if it’s too dark, short or shaky. Cancels hand shake using the still background. Measures the machine’s sound (knocks, hiss, hum at shaft speed) and its visible movement. Compares everything with the machine’s own healthy reference recordings. Replies with a voice note in English or Hindi and one clear next step: carry on, record again soon, or ask a technician. A slow-motion clip makes the machine’s movement visible, labelled as a visual aid, not a measurement. A technician summary can be forwarded to a mechanic. What makes it hard to fool. Same-spot check: notices when the phone was moved, and says how to fix it. Background-noise reference: a recording with the pump switched off stops a nearby generator being mistaken for a fault. Trend tracking: flags slow wear that no single check would catch. Repair check: shows whether the machine is back to its own normal after a repair. Built with Meta. WhatsApp Cloud API is the entire user interface. SAM 3.1 separates the machine from its surroundings. Muse Spark turns findings into natural local-language speech and reads nameplates, behind a guard that blocks invented numbers or unsafe advice. The decision itself comes from transparent signal-processing rules, never from a language model.