APESC

Analog Policy Controller for Always-On AI

APESC operates in the analog domain before digitisation, deciding when to wake AI systems and how much compute to allocate.

Unlike digital always-on processors that digitise first and decide later, APESC extracts decision-relevant features in continuous time - enabling sub-microwatt operation with superior false-wake rejection.

Two-Stage Policy Architecture

The upper domain extracts analog metrics (envelope, slope). The lower domain implements temporal voting and persistence prediction. The coupling region controls how evidence accumulates before policy states are generated. All processing occurs prior to digitisation.

Patents pending (UK)

Pre-ADC Analog Policy Extraction

APESC extracts envelope and derivative metrics from raw sensor signals before digitisation. Structured analog pathways enable relevance detection, temporal feature accumulation, and policy decisions — all in continuous time, without ADC overhead.

KEY FEATURES

  • Standalone inline placement — between producer and interconnect, transparent to existing systems

  • Sub-10ns bypass switching — prune when safe, pass through when uncertain

  • Receiver feedback loop — adapts pruning aggressiveness based on real-time quality metrics

  • Fail-open default — hardware bypass ensures data flow even if optimisation fails

  • Token pruning, not quantization — actual data removal, not precision reduction

APPLICATIONS

Domain

Use Case


Always-on AI without battery drain

Mobile


Hearing aids, smart rings, continuous health monitoring

Wearables


Wake-on-relevance for remote and embedded systems

Autonomous Sensors


Defense

Soldier systems, unattended ground sensors


COMPETITIVE POSITION

We have analysed 9 funded competitors in the always-on AI space including AONDevices, Blumind, Aspinity, Syntiant, and Innatera.

None combine pre-ADC analog processing, temporal voting, three-state output, and policy bus coordination.

Patents pending (UK)