Hardware/AISaudi Arabia2024
Flood sensor network with edge inference early-warning
Solar LoRaWAN water-level nodes that classify flash-flood risk at the edge and page authorities.
- Client
- Sadeem
- Region
- Saudi Arabia
- Sector
- Environmental / flood IoT
- Engagement
- 2024 · 18 weeks
- Team
- 2 hardware · 1 firmware · 1 ML · 1 cloud
- Status
- Deployed
The challenge
Flash floods in arid valleys rise faster than centralized monitoring can react, and cellular coverage in wadis is thin. Sadeem needed self-powered water-level and rainfall nodes that could assess flood risk locally and raise alerts over LoRaWAN in seconds.
What we built
The full pipeline, end to end — not a blurb.
- 01Hardware
Solar-powered water-level and rainfall nodes — ultrasonic and pressure sensing in a ruggedized LoRaWAN enclosure.
- 02Firmware
Sensor fusion with on-node rise-rate features and low-power scheduling.
- 03Edge inference
A flood-risk classifier on the node with tiered alert thresholds.
- 04Cloud
A LoRaWAN network server, a live risk map, and authority paging over SMS.
- 05Deployment
OTA updates and continuous network-health monitoring.
Results
23 min
flash-flood risk flagged ahead of peak on validation events
< 1 s
edge classification per reading
12 mo
field operation with zero battery swaps on solar nodes
99.8%
message delivery across 540 nodes in wadi terrain
Feeds a regional civil-defense early-warning workflow.
Have a constraint like Sadeem’s? Bring us yours.
Next engagement
Iyris
Agritech greenhouse control