Robotic olfaction lab

Smell intelligence for machines. We help robots read the chemistry of air.

Read the researchAbout the lab
VOCs & gases · on-device AI · controlled experiments
Our mission
The next generation of robots will not only see the world — they will read the chemistry of it.

The physical world is full of invisible signals. A plant under stress changes the chemistry of the air around it. A machine overheating releases different compounds. A room after a contaminating event carries a fading chemical trace.

Robots are entering these environments, but most of them are chemically blind. Aeralyte builds the research foundation for machines that can smell — controlled air sampling, compact sensor arrays, on-device AI, and rigorous experimental protocols.

Near-term: repeatable smell fingerprints. Long-term: a machine-readable language of smell that lets robots and IoT devices detect, interpret, trace, and eventually synthesize chemical signatures.

Six research pillars
We treat smell as a system, not a sensor.
Where it helps
The questions we are chasing.

Each is an open research direction — a place where reading the chemistry of air earlier, or on-device, could change what a machine can do.

01 · AgriculturePlant stress, earlierReading the chemistry of crops before disease is visible.
02 · Food qualitySpoilage signalsDetecting the volatile compounds of food as it degrades.
03 · Industrial safetyLeaks & threshold eventsSensing gases and VOCs before they become dangerous.
04 · EnvironmentAir-quality contextMonitoring particles, humidity, and chemical background.
05 · RoboticsA new sensing layerGiving mobile robots a directional sense of smell.
06 · Future healthResearch frontierA long-horizon direction we are studying carefully.
The research ladder
From near-term reality to long-term frontier.
01SenseDetect chemical changes in air
02InterpretClassify smell fingerprints
03TraceLocate sources & context
04PredictDetect risk before symptoms
05SynthesizeModel & reproduce signatures
Where we are now
Building the bench. Moving to controlled experiments.

We have designed the air-sampling protocol, a five-sensor electronic-nose array, the ESP32-S3 firmware, and the analysis pipeline that turns sensor responses into smell fingerprints. The hardware is on the bench now — we are bringing it up and beginning controlled chamber experiments.

Current stagecontrolled e-nose bench · ESP32-S3 · five-sensor arrayLive
5
sensor families
114
features per sample
5
event types
0.90
baseline accuracy