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.
We are gathering the data to train machine smell models — modern AI alongside hyperdimensional computing (HDC) — learned from controlled recordings of real air rather than simulations. Runs are recorded on a live bench under controlled, labeled conditions, with blanks and confounder controls alongside every exposure. A purpose-built sampling rig is in engineering to scale the collection. Analysis of recorded runs goes through the truth gates before any claim does.
Six research pillars:machine smell · controlled sniffing · smell fingerprints · on-device AI · drift & reality · scent synthesis
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.