Skip to content
IEMI Labs
All product designs
VisionDesign complete (stages 1–6), not built

SmokeSentinel AI

A solar daylight smoke spotter for ranches, rural communities and co-ops

Watches the horizon from a ridge or tower, spots a growing smoke column and sends your group a bearing and photo over LTE-M or satellite text, so a person can call 911 sooner.

Who buys it
Texas ranch owners, Hill Country property-owner associations and Firewise committees, and electric co-op right-of-way managers.
Why it pays
Earlier initial attack on a pasture or brush fire saves fencing, hay, cattle and structures, often tens of thousands of dollars per incident (est.). A node costs about 4–5% of one $50,000/yr agency camera station.
Target price
$2,190 Ridge (2 cameras) or $2,690 360 (4 cameras) + $39/node/month (target)
First-minute wow
A red bearing line on the map and a photo of a thin grey column 6 km out; a minute later a second node's line crosses it and shows a location to read to 911.
See the CAD designConcept page
SmokeSentinel AI concept: the idea and where each part goes
The concept. The CAD design is below, under Printable CAD.

What the research found

And how this design answers it

  • Life-safety liability → positioned as a daylight spotter that tips people who call 911; no auto-dialling, no night smoke claims, only measured range and false-alarm numbers published
  • Price no-man's-land ($4.9–6.9k + $199–349/mo) → about $1k of hardware sold at $2,190 + $39/mo for owners who will never buy a $50k/yr station
  • Separate satellite modem → one nRF9151 with LTE-M and Skylo NB-NTN fallback
  • Duty-cycled Hailo on a weak solar budget → always-on µA supervisor that powers the CM5 + Hailo-8L only in daylight or on red-flag nights; about 44 Wh/day with 11 days of autonomy
  • Clouds, dust and planned burns → persistence-and-growth rule, road masks, owner-set mute zones and a sun-glare gate
  • One camera can only give a direction → two nodes triangulate a location

Use case

Red-flag afternoon on a ranch

Ranch managerRanch handsVolunteer fire department

  1. 01Trigger: at 15:20, with humidity at 14%, a baler throws sparks into a hay field 6 km west
  2. 02Detection: in fire-weather mode the west camera sees a column grow across three frames 30 s apart
  3. 03Alert: the owner group gets bearing 268°, a cropped photo and 'about 5–7 km' over LTE-M
  4. 04Action: the manager confirms and calls 911 with the bearing; the volunteer fire department rolls with a direction from a known point

Outcome: Initial attack about 10 minutes earlier (est.); fences, hay and cattle spared

How it works

Architecture

  1. 01

    Sense: 2 or 4 fixed 12 MP Camera Module 3 Wide cameras on a 1.5-inch pipe mast, plus a humidity sensor in a radiation shield

  2. 02

    Decide: CM5 + Hailo-8L runs a smoke detector on horizon tiles every 30–60 s, with persistence and growth tracking

  3. 03

    Send: nRF9151 sends the alert and a 40 kB crop over LTE-M, or a 34-byte text over Skylo satellite

  4. 04

    Tell: push, SMS and voice to the owner group, a bearing map with triangulation, and an open CAP 1.2 feed

Hardware

Real parts, buyable today

Bill of materials: about $1,090 per Ridge pilot node including solar and battery, about $990 in production (est.).

BlockPartWhy this part
ComputeRaspberry Pi CM5 4GB on CM5 IO Board + Hailo-8L M.2Same software stack as the Edge Hub; off at night to save power
CamerasRaspberry Pi Camera Module 3 Wide (IMX708, 12 MP)45 px per degree, so a 50 m plume at 8 km is about 16 px wide
Radio + supervisorNordic nRF9151 (LooUQ MTC2-N9151 in production)LTE-M plus Skylo-certified satellite fallback and GNSS, µA sleep
Power100 W panel + Victron SmartSolar 75/15 + 12.8 V 50 Ah low-temp LiFePO4Bought, listed parts; about 6× daily need in a Texas December
EnclosurePrinted white ASA head with double-skin sun roofKeeps the head cooler in sun; no battery or mains in printed parts

Mechanical

Printable CAD

9 parts, generated from code and checked as closed solids. Pick a part to inspect it, or download the STL files to print.

See the CAD parts render
SmokeSentinel AI printable parts, rendered from the CAD
Render of the printable parts from the actual CAD. Not yet printed or photographed.

Loading the CAD…

Real CAD · drag to rotate · scroll to zoomReal CAD · swipe sideways

Open risks we are working on

  • Real detection range and false-alarm rate in Texas haze, dust and cumulus are unmeasured
  • Skylo NTN access, payload limits and pricing for a small US device are not confirmed
  • Head and battery temperatures in a Texas August may force throttling or charge cut-off
  • Carrier and PTCRB certification cost and time for a small-volume LTE-M device
  • Demand: ranches may rely on neighbours and free agency cameras instead of paying a monthly fee

Design stage only: nothing on this page is for sale. Specifications and prices are targets.