ADVENT
Applications

One instrument. Hundreds of Model Packs.

A Model Pack is software — the question the instrument answers at a given site. Entering a new area means collecting data there and training a pack, not redeveloping the machine. Six areas are in the field or in research today, and the hardware cost does not grow with the number of packs.

Diagram of one instrument running multiple Model Packs
Pack 01 · Safety

Catch the fire before there is smoke.

Overheating cable, a battery venting, a smouldering rack — chemistry announces the fault before smoke or heat does. The question is whether the instrument hears it first without crying wolf at solvent vapour.

Illustration of an e-nose watching a switchgear room and battery storage
What it watches
Electrical fire precursors, cable insulation smouldering, lithium battery and ESS off-gassing, EV charging bays
Where it sits
Switchgear rooms, cable trays, ESS enclosures, data-centre racks, underground car parks, recycling and waste facilities
Stage
Verified in-house against three background chemicals · on-device fire detection in Gyeonggi-do field trials · link to automatic fire suppression under demonstration
Boundary
A complementary channel. It does not replace statutory fire protection systems or a battery management system.
Pack 02 · Environment & odour

Turn an odour complaint into a measurement.

A complaint arrives as a description. Mk.2E turns it into data: which class of odour, from where, and when — continuously, instead of by a panel after the fact.

Illustration of odour monitoring at a wastewater facility and residential boundary
What it watches
Chemical, wastewater and sewage odour classes; emissions and complaint monitoring around industrial complexes and residential areas
Where it sits
Wastewater and sewage facilities, industrial complex boundaries, landfills and incinerators, manholes
Stage
Blind class separation verified · residential odour source tracking with Korea Environment Corporation (October 2025)
Boundary
Not a legally mandated measuring instrument. It complements statutory measurement, it does not replace it.
Pack 03 · Industrial & food quality

Is today’s batch the batch the customer approved?

Sensory panels are slow, expensive and vary from person to person. A fingerprint compares today’s batch against a reference, not against someone’s memory of last month.

Illustration of batch comparison on a food and beverage line
What it watches
Batch-to-batch comparison, process drift, raw-material and finished-product checks on the line
Where it sits
Food and beverage lines, fermentation, fragrance and chemical processing
Stage
In discussion and trials with industrial and food-processing operators
Boundary
Targets are confirmed by collecting data on the line before a pack is trained.
Pack 04 · Livestock & agriculture

Read the air in the barn.

Ammonia and odour load affect animal health, worker safety and the neighbours. The same instrument also watches for fire precursors in housing that is hard to supervise at night.

Illustration of air monitoring in a livestock barn
What it watches
Ammonia and hydrogen sulfide load, odour emissions, early fire precursors, early warning signs in animal housing
Where it sits
Poultry and pig houses, manure handling, storage facilities
Stage
Proposals and pilot design with livestock and agricultural partners
Boundary
Not a veterinary diagnostic. Signals support management decisions on site.
Pack 05 · Healthcare & care

A research question, treated as one.

Do scent signatures carry information useful in screening and care settings? We are studying it — carefully, and without claims we cannot back.

Illustration of scent-signature research in a care setting
What is being studied
Scent-signature research for screening and care environments, including odour monitoring in care facilities
Stage
Research stage
Boundary
GraphAISense is not a medical device and makes no diagnostic claim.
Research stage — not a diagnostic product
Pack 06 · Robotics & autonomous systems

Give a patrol robot a sense of smell.

A mobile platform can see and hear, but it cannot tell a leak, a spill or a burning cable apart by chemistry. An e-nose module gives it a chemical sense of place.

Illustration of a patrol robot carrying an e-nose module
What it adds
Chemical detection and material identification on mobile and autonomous platforms, covering blind spots that fixed units miss
Why on-device matters
Recognition completes on the module in under a second, so the robot acts without waiting on a network
Stage
E-nose on a SEO Robotics patrol robot at the Incheon Port passenger terminal · collaboration test with Korea’s National Fire Agency on Hyundai MobED
Boundary
Integration details are scoped per platform; robot data links over MQTT.
A seventh pack

If your question is not on this list, it is a data question, not a hardware question.

A new area means measuring under known conditions, labelling against real events and training a pack. Tell us what you need to detect, what it has to be told apart from, and what decision the answer must support — we would rather start from the chemistry than from a brochure.