AI-powered acoustic sensing network
Solar-powered sensor nodes around fields, sheepfolds and facilities tell humans, vehicles, wild boars, wolves, bears, herds and drones apart — right on the device — locate them and alert you instantly.
Classifies on the device
01 The problem
Farmers guard their fields, shepherds their flocks and operators their sites by keeping watch at night. It is expensive, exhausting and unsustainable — and nothing records what actually happened.
The countryside lacks a sensing layer that is cheap, widespread, off-grid and able to tell what is approaching. Şahit builds that layer.
Sources: TARSİM, TÜİK (June 2026), local and national press reports.
02 How it works
Every node processes sound and motion on its own; only the event leaves the device. When several nodes hear the same event, the source is located.
The node listens continuously at very low power with a four-microphone array and a motion sensor (PIR). An ultra-low-power event detector is always on.
When something happens, the on-device AI model wakes up and classifies the sound: human, vehicle, wild boar, wolf/dog, bear, livestock herd or drone.
Neighbouring nodes confirm the event. The source is located from the differences in sound arrival times. No high-priority alert is sent unless at least two nodes heard it.
An instant alert goes to your phone. For animals, a species-specific light and sound deterrent is triggered — with random patterns so they do not get used to it.
Every event is logged with time, place and class. The event timeline becomes a damage-evidence report that supports insurance assessment.
03 What it hears
Every sound leaves its own trace. The on-device model tells these traces apart and is trained with field recordings to separate the familiar sounds of the countryside from real threats.
Corn, sugar beet, strawberries, vegetables and hazelnuts
ActionDeterrent + alert + damage record
Around fields, folds and sites at night
ActionInstant alert · raw data never stored
Vehicles approaching at night, roadside
ActionApproach alarm
Sheepfolds and pastures
ActionDeterrent + alert
Fields and folds at the forest edge
ActionDeterrent + alert
Passing herds or your own
ActionLogged only · no alarm
Around sites, low altitude
ActionEarly warning (Şahit Perimeter)
Learned from labelled field recordings — to keep false alarms to a minimum.
Target: 7 classes in the commercial version. The first field prototype prioritises wild boar, human and vehicle.
04 Technology
Designed, programmed, trained, assembled and tested in Türkiye. Every node makes its own energy with a solar panel and a LiFePO4 battery — no cables, sockets or Wi-Fi needed.
10 W panel, sized to cover daily consumption even in winter sun
4 digital MEMS microphones; direction finding inside the node
PIR; ultra-low-power wake-up trigger
Microcontroller with a neural accelerator; the model runs on the device
Long-range, low-power link from node to gateway
~75 Wh with low-temperature charge protection
Dust- and water-proof, with a tamper sensor
Thermal camera or 60 GHz mmWave radar
The design that sets the average power: an ultra-low-power event detector is always on, the classifier only wakes up when something happens.
Target with design margin: < 0.4 W
05 Data path
Sound is processed on the device. Nodes connect to a gateway over LoRa, the gateway to the platform over 4G. The platform locates the source, manages the fleet and remote updates, and the alert lands on your phone.
{
"class": "wild_boar",
"confidence": 0.94,
"confirming_nodes": 3,
"location_error_m": 12,
"time": "2026-10-07T02:14:07+03:00",
"raw_audio": null
}Only what, where and when reaches the platform.
No raw data is kept for human detections — privacy by design.
Fleet health, battery and signal monitoring; models updated in the field.
06 Use cases
The same Şahit node and model work in fields, sheepfolds and around critical sites on one platform.
Farmers, cooperatives, municipalities, insurers
Wild boar detection, deterrent triggering and damage-evidence reports for insurance assessment.
Sample setup
Sample alert
Sheep, goat and cattle farms
Wolf and thief alerts, night-time vehicle approach alarms. Far more economical than a collar per animal for small flocks.
Sample setup
Sample alert
Critical site operators
Perimeter security and low-altitude drone early warning for solar and wind farms, substations, dams and pipeline stations.
Sample setup
Sample alert
07 Targets
Şahit is developed by measuring in the field. These are the targets set for the first field prototype and the commercial version.
These are targets. They will be measured in field pilots against neighbouring control areas without nodes.
08 Roadmap
Sensing chain on off-the-shelf development boards; class separation and power measurements in the lab.
Our own board, IP67 enclosure and solar panel; the first field nodes.
One season in three different climates and terrains; damage compared with control areas.
CE pre-tests, gateway, deterrent and localisation.
First commercial batch and CE certification.
09 Principles
Sound is processed on the device; no raw data is kept for humans; information signs on site.
Light and sound only. Non-lethal, with random patterns animals do not get used to.
Solar panel and LiFePO4 battery; no cables or sockets.
Design, software, model, assembly and testing in Türkiye.
10 FAQ
No. Şahit uses sound and motion; it needs no line of sight and works at night and inside crops. Where needed, a thermal camera or radar module can be added.
Nodes are off-grid: they run on a solar panel and a battery. Nodes connect to a gateway over LoRa and the gateway to the platform over 4G — no Wi-Fi needed on site.
No. Sound is processed on the device and only the event is sent. No raw audio is kept for human detections.
No. Deterrents use only light and sound. Patterns change randomly, so animals do not get used to them.
No high-priority alert is sent unless at least two nodes heard the event. Low-confidence events are flagged for review. You can mark every alert as right or wrong in the app — that feedback keeps improving the model.
The time, place and type of each event are recorded. That timeline becomes a damage-evidence report that supports insurance assessment.
Şahit is currently at the prototype stage. Write to us to take part in field pilots or to collaborate.
Contact
Are you a cooperative, municipality, farm, insurer or site operator? Write to us about Şahit field pilots and collaborations.
Şahit is developed by HB Tech.Visit hbtech.com.tr