FOR IOT AND EMBEDDED DEVICES

Sensor readings that join the data lake.

A photo tells you what is on the wall; readings from inside the building tell you why. Mould Detect's prototype board already posts into the same data lake as every scan and weather feed, on a schema where a new sensor adds a key, never a new pipeline.

THE HARDWARE

A board running today, and a diffuser in design.

Left, a recording of the test board feeding the live dashboard. Right, the concept design of the sensing diffuser, numbered part by part.

The ESP8266 test board, on a battery pack, feeding the live dashboard on a laptop
PrototypeAn ESP8266 Witty board posts batched HTTPS readings about every 30 seconds in development. Its one light sensor is there to prove the pipeline; the physics comes next.
Concept design of the Mould Detect ultrasonic oil diffuser, drawn as an isometric cutaway on a blue blueprint sheet, titled Sensing and response in one device and marked In R&D, 2027 prototype. Nine numbered parts run from top to bottom: top cover, an air outlet for diffused oil droplets; a 5 V cooling fan; a 2.4 MHz ultrasonic piezoelectric transducer; a microcontroller board that controls timing, duration and dwell rate; a sensor board for temperature, humidity, moisture, formaldehyde, CO2, UV light and VOCs; an oil reservoir for a plant and tree oil blend; the housing; a rechargeable lithium-ion battery; and a 5 V USB power supply. A detail circle shows the 2.4 MHz piezoelectric disc.
Planned 2027The sensing diffuser, from the Ventures Australia 2026 deck: ultrasonic diffusion plus a sensor board for temperature, humidity, moisture, formaldehyde, carbon dioxide, UV and VOCs. A concept in R&D, planned as a 2027 prototype. Concept design by Rozario Chivers, Vladimir Lekic and Lucas Chivers.

BUILD PHASES

What each phase adds, and where it stands.

Read top to bottom. Every phase posts into the same schema, so nothing downstream changes as sensors are added.

PhaseStatusWhat it adds
1 to 2bDoneA local prototype, cloud ingest into the data lake, and a hosted read-only dashboard.
3Next buildTemperature, relative humidity, dew point and pressure (SHT31 or BME280), and a calibrated light sensor.
4RoadmapOn and beside the diffuser: VOC (SGP40), carbon dioxide (SCD40), formaldehyde, UV, surface moisture, water depth and water leak, plus the diffuser's own on/off and duty state.

Particulate sensing is not planned: outdoor particulates already come from CAMS. Water depth and leak sensing matter after floods and plumbing failures, which raise indoor humidity.

MOULD DETECT BUSINESS PLAN v3.1 · 14 OCT 2026

THE PIPELINE

Cheap to run, and the same schema for every sensor.

Three design decisions an engineer will ask about first.

  • Device, time, key, value.

    Readings are stored as four fields, so a new sensor adds a key rather than a table, a stream or a service.

  • Serverless ingest.

    Device to a Lambda function URL, into the S3 lake partitioned by device and day, rolled up and served through a read API to the dashboard.

  • About 2.3 US cents per device a month.

    Modelled running cost at the 15-minute production cadence. Idle costs nothing, because nothing is always on.

WHERE READINGS GO

Indoor readings teach the Atlas what outdoor data cannot.

Sources on the left feed the lake; the lake feeds the Mould Atlas. Sensor boards are how phase 2, indoor calibration, will be learned.

  1. Independent environmental feeds.LiveERA5-Land climate history, NOAA GFS forecasts, GloFAS river flood and CAMS air quality, collected for Australia today.
  2. Readings from inside buildings.PrototypeSensor boards post to the same lake on a sensor-agnostic schema: a new sensor adds a key, not a new pipeline.
  3. Scans, assessments and lab results.CollectingAI Vision scans, Mould Assessment reports and partner lab results become the labels the Atlas learns from.
  4. Phases 0 and 1 run today.LiveThe data lake, and a published VTT mould-growth model per location served as /v1/mould-index. It is driven by outdoor conditions and has not yet been validated against indoor damage.
  5. Phases 2 to 4 are the roadmap.RoadmapLearn the outdoor-to-indoor transfer by building type, train on verified outcomes, then predict risk by building type, postcode, climate zone and season.

FROM READING TO RESPONSE

A reading that crosses a threshold can act.

Read top to bottom: data sources, a threshold agreed in advance, and the responses that fire.

  1. Oracles.LiveIndependent data no party to the contract controls: weather and flood feeds, in-building sensors, photo scans, specialist assessments and lab results.
  2. A threshold agreed in advance.IllustrativeFor example relative humidity above X% for N days, a GloFAS flood exceedance, or a condensation risk from the dew-point spread.
  3. A. Environmental control.Phase 4Ventilation, a dehumidifier or the diffuser switched on when a trigger fires.
  4. B. Alerts.PrototypeIoT alerting to landlords, facilities and maintenance teams, early.
  5. C. Remediation booking.PartnersA lab test or an approved remediator, booked through the supplier list.
  6. D. Data and API feed.LiveThe same readings delivered to partners, including insurers, into their own systems.
The technical pattern only. No insurer pilot or contract exists today.

Build sensors with us.

If you design boards, firmware or building sensors, tell us what you measure. The schema is ready for a new key.

Email usSee the live dashboard