Proof & Validation

What we can demonstrate today, and what we won’t claim yet. Additional test artifacts available under NDA.

Request Evaluation Packet
Request Technical Appendix (NDA)

Validation Sequence

We sequence validation to build confidence:

Technical → Operational → Institutional.

Each layer includes explicit status and supporting evidence.

Status Definitions

Validated: Validated with documented results and an evaluator-ready artifact.

In Validation: Under active testing with defined acceptance criteria.

Not Claimed: Not claimed publicly at this time.

Technical Proof

What the platform can demonstrate under controlled conditions.

  • Claim
    The system can accurately detect human screaming sounds in real time at distances of up to 50 meters, with detection accuracy above 98.7%.

    Test Conditions
    Indoor environments; mixed background noise (TV, people talking, voices); single and multiple sound sources; day and night testing.

    Status
    Validated

    Evidence Artifact
    Field test recordings and detection logs (available under NDA)

  • Claim
    The system can reliably detect human coughing sounds and distinguish them from similar sounds (throat clearing, speech, background noise) with over 96.8% accuracy.

    Test Conditions
    Indoor and outdoor environments; varying microphone sensitivity levels; different cough intensities; mixed ambient noise.

    Status
    Validated

    Evidence Artifact
    Annotated audio datasets and validation reports (available under NDA)

  • Claim
    The system can detect and classify gunshot sounds in real time and differentiate them from acoustically similar impulsive sounds (fireworks, car backfires) with over 98.2% accuracy.

    Test Conditions
    Indoor environments (via portable speaker); controlled test recordings and real-world datasets; varying distances and sound pressure levels.

    Status
    Validated (controlled datasets) / Field validation in progress (outdoor live-fire)

    Evidence Artifact
    Controlled test data, classification metrics, and field validation logs (available under NDA)

  • Claim
    The system can estimate the direction of arrival of gunshot sounds with an angular accuracy of ±20° using a multi-microphone configuration.

    Test Conditions
    Quad-microphone array; outdoor testing; fixed and moving sound sources; variable environmental conditions.

    Status
    In validation (targeting completion in 2–3 weeks)

    Evidence Artifact
    Prototype DoA visualizations and test datasets (available under NDA)

Operational Proof

What the platform demonstrates in real environments with real constraints.

  • Claim
    The system can run continuously on the edge for extended periods without human intervention, maintaining stable system operation and alerting behavior.

    Operational Context
    Remote outdoor deployment; device mounted on fixed infrastructure; battery-powered; intermittent connectivity; multi-week runtime.

    What We Logged (Operational Metrics)
    Uptime; event counts; health checks; power trends; fault/reboot logs.

    Status
    Proven (Operational)

    Last updated
    Nov 2025 (ongoing)

  • Claim
    The system maintains controlled false-positive performance in real-world environments with high ambient noise and variability.

    Operational Context
    Outdoor environments including wind/rain, background human activity, and environmental noise; continuous monitoring with static placement.

    What We Logged (Operational Metrics)
    True/false event tags; alert frequency; conditions at detection time; operator feedback tags.

    Status
    Proven (Operational)

    Last updated
    Dec 2025 (ongoing)

  • Claim
    The system delivers alerts and logs from remote locations to central dashboards over low-bandwidth links with consistent delivery performance.

    Operational Context
    Remote sites with limited connectivity; edge processing enabled; alerts transmitted via low-bandwidth channels (e.g., LoRa/LoRaWAN) with fallback where available.

    What We Logged (Operational Metrics)
    Delivery time; message success rate; connectivity uptime; packet size; retry/fallback events.

    Status
    Proven (Operational) (or Validated if formal criteria/3rd-party)

    Last updated
    Late 2025

Institutional Validation

Where the platform is evaluated and how it moves into formal programs.

Engagement Pathways: We support structured evaluations through appropriate government and partner channels.

Validation Flow: Technical proof first, then operational evaluation, then formal adoption when criteria are met.

Current Posture: We prioritize qualified evaluations and controlled pilots over broad public claims.

Some evaluation activity is not publicly disclosed. Details and supporting artifacts are available under NDA.

Capability Registry: What It Can Hear

Updated only when a new sound class clears a proof gate and has a publishable artifact.

The registry is a living index, not an open-ended claims list.

  • Sound Class: Human screaming sounds (non-speech distress)

    Output Type: Detection event + confidence score (with event label)

    Status: Validated

    Artifact Available: NDA (field test recordings + detection logs)

    Last Updated: Jan 2026

  • Sound Class: Human coughing sounds (non-speech)

    Output Type: Detection event + confidence score (with event label)

    Status: Validated

    Artifact Available: NDA (annotated datasets + validation report)

    Last Updated: Jan 2026

  • Sound Class: Gunshot sounds (impulsive acoustic events)

    Output Type: Detection + classification label + confidence score

    Status: Validated (indoor + recorded datasets); Outdoor live-fire validation in progress

    Artifact Available: NDA (controlled test data + classification metrics + validation logs)

    Last Updated: Jan 2026

  • Sound Class: Gunshot sounds (multi-microphone configuration)

    Output Type: Directional estimate (DoA) + confidence score

    Status: In Validation (multi-mic array)

    Artifact Available: NDA (prototype DoA visualizations + test datasets)

    Last Updated: Jan 2026

Claims Discipline

  • We do not publish weekly capability expansion without validation.

  • We separate Proven vs In Validation vs Not Claimed.

  • We prefer precision over breadth.

  • We share deeper technical detail under NDA.

Evaluator Resources

Download Evaluator One-Pager
Request Technical Appendix (NDA)
Download Demo Protocol Overview
Download Integration Pathway Outline

Request an Evaluation Brief

If you are assessing SGI for validation, integration, or acquisition pathways, request the evaluation brief and we will route you appropriately.

NDA available upon request for technical materials.