Alexander Mecca.
Project directory

01Personal Products and Experiments

SENTINEL Recognition and Presence

Actual archived product · Privacy-safe capture

An archived local-AI system that turned discrete camera observations into reviewable identity decisions and room-level last-seen state. Visual description stayed separate from deterministic identity matching; low-confidence or unknown observations remained human-review evidence instead of trusted presence. The camera pipeline was retired in August 2026, and its work-activity subsystem evolved into CHRONOS.

Computer visionPresenceLocal AIReal-time UI

The discrete-observation model

Observe locally. Decide visibly. Update room-level state only when evidence arrives.

  1. 01

    Observe

    Event-driven scanning produces a discrete camera observation rather than pretending the system knows a continuous path.

  2. 02

    Describe + Match

    Local visual description and deterministic identity matching remain separate, inspectable signals with visible confidence.

  3. 03

    Update Last Seen

    A trusted recognition updates the person’s last-seen room and event timeline.

  4. 04

    Review

    Unknown or low-confidence observations remain evidence for human review; health telemetry keeps the quality of the pipeline observable.

Actual SENTINEL product interface · Separate synthetic database and fictional floor plan · Real room-level state rendering · Private names, home layout, cameras, images, and device identifiers omitted or redacted

What the accomplishment is

The implemented capability—not a substitute illustration.

An archived local-AI system that turned discrete camera observations into reviewable identity decisions and room-level last-seen state. Visual description stayed separate from deterministic identity matching; low-confidence or unknown observations remained human-review evidence instead of trusted presence. The camera pipeline was retired in August 2026, and its work-activity subsystem evolved into CHRONOS.