Mini-Model · Edge / On-Device Ontologies
Mini-Model is the reference vocabulary for AI operating under hard resource constraints. Where macro and micro models describe what intelligence does, mini-model describes what intelligence must not cost — in memory, power, latency, and connectivity.
Edge Use Cases
Each scenario defines a distinct constraint envelope. A model that passes all three is genuinely edge-deployable across the IoT, mobile, and embedded spectrum.
Keyword classification and anomaly detection running at microwatt power budgets on embedded MCUs — no cloud round-trip, no persistent memory.
On-device NLP for privacy-preserving UX — autocomplete, intent classification, and content moderation without leaving the device.
Real-time control inference in automotive ECUs, industrial PLCs, and avionics — deterministic latency under hard real-time constraints.
Story Figures
Three figures that show how a compact model earns its place: the constraint envelope it must fit, the people who share the deployment decision, and the comparison that turns a task into a concrete choice.
