+ Technology · The four principles
The platform underneath everything.
Four engineering principles that define an Ambient IoT device — and the patent portfolio that protects them.
Four engineering principles that define an Ambient IoT device — and the patent portfolio that protects them.
An Ambient IoT device pulls its power out of the room it's in. Stray RF from nearby radios, indoor fluorescent or LED light, vibration from machinery, thermal gradients across a refrigerated wall — all of it can be captured, rectified, and stored on a printed micro-supercap or paper-thin battery.
The duty cycle is brutal: most of the device's life is spent sleeping at nanoamps. It wakes for milliseconds, takes a reading, transmits, and returns to sleep before the harvested charge runs out.
The result: a device that runs for the lifetime of the asset it's stuck to, with zero maintenance.
A traditional IoT device assumes infrastructure: a cellular tower, a LoRa gateway, a Wi-Fi access point with a known SSID. That assumption fails on a ship, in a cool room, in a fab cleanroom, on a remote farm.
Ambient IoT devices ride whatever's nearby. A passing smartphone with the right app. A worker's BLE (Bluetooth Low Energy) wearable. A satellite uplink (Iridium, Kinéis) when nothing terrestrial is in range. A consumer mesh like Amazon Sidewalk in dense urban areas.
The cloud reassembles the partial messages. The device doesn't care which radio carried which packet.
If a device costs $20, you put it on a pallet. If it costs 20¢, you put it on every carton. If it costs 2¢, you put it inside the label.
We pursue cost relentlessly: printed antennas and circuitry, off-the-shelf MCUs in cheap form factors, sensors integrated into the PCBA (printed-circuit board assembly) rather than bolted on, packaging that is the device. Where appropriate, the device is single-use and compostable — the journey ends, the label biodegrades, the data lives on in the cloud.
That economics inversion is what unlocks tracking every item rather than just every shipment.
Cloud-based inference assumes connectivity. Ambient IoT can't. So the model lives on the device.
A TinyML model — on-device machine learning running on the MCU, co-located on the same PCBA as the sensor, fingerprints the raw signal, compensates for drift, classifies events, and decides what's worth sending. Most of what the sensor reads is uneventful. The radio only fires when something matters.
That single architectural choice cascades through the platform: lower power, lower bandwidth, lower cost per data point, lower false-alarm rate, and a real-time decision in milliseconds rather than minutes.
When a device costs cents and lasts forever, the business case changes. You stop tracking shipments and start tracking items. You stop sampling environments and start measuring them continuously. You stop logging data and start predicting from it.
Tony Raftis is the founder and CEO of Ambient IoT Pty Ltd, the Australian deep-tech company developing battery-free physical-world sensing under the StiknTrak and Moseley brands. His 40-plus year career spans IBM Australia and the ASX-listed Volante Computers — alongside ventures in WWII aircraft restoration (Platinum Fighter Sales) and motorsport, including the 2008 World GT Championship in Bahrain. He is the sole inventor on every Ambient IoT patent application currently pending, and a multiple patent holder in biometrics and sensing technologies. Based in Queensland, Australia. LinkedIn →
Ambient IoT licenses platform IP, partners on bespoke applications, and supplies its brands at scale. We're always interested in conversations with logistics, pharma, semiconductor and agritech operators.