Teaching machines the tacit knowledge of the senses.
OFLab builds HSTKG, a cross-modal AI that turns the non-verbal feel of sound, scent, and emotion into machine-representable knowledge.
A research lab bridging foundational science and commercial AI.
Today's AI models reason fluently over language, yet the tacit, sensory knowledge beneath human experience stays out of reach. OFLab researches that foundation: the cross-modal structure of the senses. From a quiet base in Jeju, we turn this foundational research into capabilities that strengthen commercial AI models.
Foundational research
We study the non-verbal knowledge of the senses, the layer below language that today's models cannot reach.
Bridge to commercial AI
That research feeds back as capabilities (embeddings, ontologies, and reasoning) that help applied AI models.
Built in Jeju
A deep-tech lab grounded in the calm and nature of Jeju Island: focus over noise.
Hyper-Sensory Tacit Knowledge Graph
HSTKG aligns heterogeneous sensory modalities through contrastive learning, turning the tacit knowledge of the senses into a machine-representable, reasoning-ready form, for example, inferring a related scent or a coined word from the mood a given timbre evokes.
Aligned by contrastive learning into a shared cross-modal space, so meaning can move between senses that language never bridged.
- VECTOR DBMilvus / Pinecone: high-dimensional sensory embeddings
- GRAPH DBNeo4j: semantic relations between senses
- REASONINGGraphRAG engine: cross-modal inference
- SCALE · AWSGPU · vector stores · serverless: multimodal training & inference
Studying the senses, from a land made of them.
HSTKG studies what the senses know before words: sound, scent, emotion. On Jeju, those aren't abstractions in a dataset; they're the daily environment: wind moving through the oreum and the sea, citrus groves and volcanic earth, light and mist that shift by the hour. We build a science of the senses in a place that is unusually rich in them.
The material, lived daily
A landscape thick with sound, scent, and mood, the very raw material of our research, all around us.
Stillness for deep work
Foundational research needs long, quiet attention. Away from the hub's churn, Jeju gives the stillness deep work demands.
A choice, not an outpost
A research culture rooted in nature and slowness, one that draws people who want depth over hustle.
The bigger picture
A deeper overview of the HSTKG research program is in preparation.
Get in touch
For research, partnership, or general inquiries, we'd be glad to hear from you.
[email protected]