
AirLeak AI
A wall or pole-mounted ultrasonic sensor disc with a camera that listens for compressed-air and gas leaks.
- Hears leaks around the clock
- Marks leaks on a camera image
- Estimates each leak's yearly cost
From $1,490 per unit
DetailsDust-plume camera and PM4 trend that flags failed silica controls
Software: $29 per month per unit
Concept, not for sale yet. Tell us if you want it; that decides what we build.
Rate this idea and tell us what to change
Drag to rotate · right-drag to move · scroll to zoom · click a part for its size and placeSwipe to rotate · pinch to zoom · two fingers to move · tap a part for its size
How it fits together
Scroll or tap a part. Press “See the 3D CAD design” to watch the model, built from our CAD, light each part up and show how it connects.

Real-time 3D render · drag to rotate · final design may varySwipe sideways to rotate
01
The whole DustWatch AI
A box on each saw checks water flows while the tool runs. A tower light warns on a dry cut; a dust monitor reads air.
02Brain
Adafruit ESP32-S3 Feather
The brain: it runs the wetcheck station firmware program
03
Hall flow sensor
Counts water pulses on the saw's supply hose
Adafruit ESP32-S3 Feather · GPIO pin (pulse count) · Flow pulses

04
Tool-on sensor
Tells the box the saw motor is running
Adafruit ESP32-S3 Feather · Dry contact, or Wi-Fi · Tool on or off
05
Dust monitor (SPS30)
Measures fine dust in the shop air once a minute
ESP32-S3 Feather · Wi-Fi (MQTT) · PM1 to PM10, humidity

06
Tower light and buzzer
Green wet, yellow warn, red and buzzer when dry
Adafruit ESP32-S3 Feather · 4 GPIO pins (FETs) · Lamp and buzzer states
07
Mute button
Silences the buzzer for 60 s; each press is logged
Adafruit ESP32-S3 Feather · GPIO pin · Button press
08
5 V step-down regulator
Powers the Feather from the 12 V adapter
Adafruit ESP32-S3 Feather · 5 V power lead · 5 V, 1 A
09
All connected
7 key parts, one program, and your phone
What it is
A bench-mounted camera and dust sensor that warns when dust controls on cutting tasks fail.
What it does
Who it's for
Stone fabricators, contractors and manufacturers with silica-generating tasks.
The idea
The problem today
Small countertop and stone shops cut engineered stone that throws fine silica dust when it is cut dry. A closed valve, a kinked hose or a quick dry cut at the polisher goes unseen, and the shop has no record to show its insurer or hygienist that it cuts wet. Personal air samples are costly, slow and only cover a day or two.
With DustWatch AI
A station box beside each saw, grinder or polisher knows when the tool is running and when the water is flowing, and lights yellow then red with a buzzer within seconds of a dry run. One dust monitor on the shop wall shows dust events as a trend. On the 1st of each month the owner approves a wet-cutting report for the insurer or hygienist. It never gives a silica level or an exposure figure.
How the parts connect
Every part, what links it to the brain, and how the data reaches your phone. A solid line is a cable; a dashed line is a radio link.
The program
for (;;) { // 50 ms tick
bool tool_on = shelly_current_a() > I_ON || gpio_get_level(CURRENT_SW) == 0;
float flow_lpm = pcnt_pulses_last_1s() / K_PULSE_PER_LPM;
if (tool_on && flow_lpm < Q_MIN_LPM) low_ms += 50;
else if (flow_lpm >= Q_MIN_LPM) low_ms = 0;
if (!tool_on) state = IDLE;
else if (low_ms > 5000) state = DRY; // red + buzzer
else if (low_ms > 2000) state = WARN; // yellow + chirp
else state = RUN_WET; // green
beacon_set(state);
if (state_changed()) mqtt_publish_event(state, hash_chain());
vTaskDelay(pdMS_TO_TICKS(50));
}To your phone
Owner phone (web app), WhatsApp or SMS
One screen for each job DustWatch AI does.
Dry cutA cutter starts the bridge saw with the water valve shut. Within 2 seconds the tower light turns red and the buzzer sounds.
Wet-cut recordEvery minute the saw runs is logged as wet or dry, so the owner can show the shop cuts wet.
Shop dustAfter the shift someone dry-sweeps the floor. The dust monitor sees the trend jump and flags it. It is a trend, not exposure.
Monthly reportAt month end the app drafts a report of wet-cut time and dry events. It goes to the insurer only after the owner approves.
App screens with sample data. The interface is in development.
| Dry-cut event | Sierra 1, 41 s dry | Each event |
|---|---|---|
| Wet share | 98% of tool time | Every 60 s |
| Flow | 6.1 L/min | Every 60 s |
| Dust level | PM2.5 9.8 µg/m³ | Every 60 s |
| Monthly report | Draft PDF to approve | Monthly |
Alarms work with no internet. Reports go to the insurer only after the owner approves them.
CAD design · Design complete (stages 1–6, pilot rev A), not built, not validated

| AI processor | Kendryte K230D (on-device plume detection) |
|---|---|
| Camera | Sony IMX462 low-light camera |
| Sensors | Sensirion SEN66 / SPS30-class optical particulate sensor (PM1 / PM2.5 / PM4 / PM10) · Temperature and humidity |
| Connectivity | Wi-Fi · Optional LTE · Bluetooth for setup |
| Power | Mains via USB-C adapter or 12–24 V DC; optional 8-hour battery for moving between benches (target) |
| Enclosure | Rounded light-grey box with a teal accent ring, a downward-facing Ø3 cm protected inlet cowl and a camera eye in the upper front corner; magnetic base or clamp |
| Dimensions | 20 W × 14 H × 10 D cm; mounted at 1.2–1.8 m on a bench or column |
| Weight | ~1.2 kg (est.) |
| Ingress rating | IP54 with heated, dust-filtered sample inlet (target) |
| Particulate | Optical PM1 / PM2.5 / PM4 / PM10 trend |
|---|---|
| What it measures | Total particulate by light scattering; not crystalline silica |
| Camera | Low-light plume and tool detection |
| Site correlation | Optional factor entered by your hygienist from lab samples |
| Alerts | Plume, suspected dry cut, PM4 spike, rising trend |
|---|---|
| Channels | App, email, SMS, Teams / Slack |
| Reports | Shift dust-trend reports by bench and task |
| Data out | Counts, trends and blurred event stills |
| Power | Mains / USB-C or 12–24 V DC; 8 h battery option |
|---|---|
| Size | 20 × 14 × 10 cm |
| Ingress | IP54 with heated inlet (target) |
| Mounting | Magnetic base or clamp, 1.2–1.8 m |
Targets for production units; they may change during development.
Hardware: $799 per unit
Software: $29 per month per unit
Target pricing. Fabricator 3-pack at $2,190 + $69/mo. Lab sampling for exposure assessment is not included and is still needed.
$799 + $29/mo per unit
One workstation with plume and dust-trend alerts.
$2,190 + $69/mo
Three units for a typical stone shop or cutting area.
Custom
For contractors and manufacturers with many sites.
Indicative launch pricing in USD, excluding taxes and shipping. Final pricing is confirmed before any balance is charged.
Community
DustWatch AI is in development. Rate the idea, tell us what to change about the features, specs or price, and we'll build it with your feedback.
How much people want it
–
No ratings yet
Follow DustWatch AI
Get build updates and first access when pre-orders open.
No comments yet. Tell us what you'd want from DustWatch AI.
Questions
Tell us what this would save you, or ask anything. We reply within two business days.
A member of the IEMI Labs team will reply within two business days.
From the product lab · ideation phase
No. Optical sensors measure total particulate by light scattering and cannot measure respirable crystalline silica. DustWatch shows a dust trend and flags control failures. It does not show silica concentrations or exposure limits, and exposure assessment still needs personal sampling with lab analysis by a competent person.
The camera points at the workstation, and people are blurred on the device. Only counts, trends and blurred event stills leave the unit. There is no identity, face recognition or emotion and fatigue inference.
No. It is in development and holds no certifications. It is an early-warning aid, not a safety-rated system or a measuring instrument, and it does not make you compliant. It helps you evidence that controls are used. It is not intrinsically safe and is not for hazardous (classified) areas. CE/UKCA/FCC certification testing is planned.
Fix it to a bench or column at 1.2–1.8 m with the magnetic base or clamp, point the camera at the workstation and plug it in. Setup in the app is designed to take a few minutes per unit.
Wi-Fi, or the optional LTE module for sites without a network. Alerts and plume detection run on the device, and events are sent when a connection is available.
No. It is in development and the specifications are targets. It is a concept: we are not taking orders, reservations or deposits, and it has no ship date yet.
We are planning a 2-year limited hardware warranty covering manufacturing defects. Final terms will be published before shipping begins.

A wall or pole-mounted ultrasonic sensor disc with a camera that listens for compressed-air and gas leaks.
From $1,490 per unit
Details
A small battery sensor that screws onto a motor or pump and spots developing faults.
From $249 per node (HD variant $349); gateway $399
Details
A one-per-machine node with one brain that does five monitoring jobs at a compressor or other big machine.
From $1,190 per machine (4 joint tags included); site LoRa gateway $399 once
DetailsDustWatch AI
Concept · target $799