Skip to content
All work

Hardware/AIPakistan2023

Health & heat-detection collars for dairy herds

Neck-motion sensing and on-device ML that flags estrus and early illness before a vet ever visits.

Client
Cowlar
Region
Pakistan
Sector
Livestock health IoT
Engagement
2023 · 20 weeks
Team
2 firmware · 1 hardware · 1 ML · 1 cloud
Status
Deployed

The challenge

Dairy farms miss estrus windows and early illness because manual observation is sparse, and a missed heat cycle costs a full breeding month. Cowlar needed a rugged, low-power collar that could infer health and heat events on the animal and still sync over patchy rural connectivity.

What we built

The full pipeline, end to end — not a blurb.

  1. 01Hardware

    A low-power collar PCB — accelerometer and temperature sensing, an nRF52 SoC, and a LoRa radio sized for month-scale battery life.

  2. 02Firmware

    On-device motion feature extraction and activity classification with adaptive duty cycling, so only events leave the collar.

  3. 03Edge ML

    A rumination and estrus classifier quantized to run inside the MCU budget.

  4. 04Cloud

    LoRaWAN gateways into an ingest pipeline with per-animal baselines and farmer SMS/app alerts.

  5. 05Deployment

    OTA firmware and a fleet-health dashboard across herds.

Results

92%

estrus-detection recall against vet-confirmed events

9 h

earlier heat flagging than visual observation, on average

14 mo

collar battery life at a 30-second sensing cadence

99.6%

telemetry delivery across 6,800 collars on rural LoRa

Now the farm’s primary heat-detection signal across commercial dairy sites.

Have a constraint like Cowlar’s? Bring us yours.

Start a project

Next engagement

Trukkr

Fleet telematics