Companion Health · Platform Document · v1.0
The future of pet care is continuous intelligence.
Companion Health is biological intelligence infrastructure for the global pet ecosystem — operating quietly, continuously, beneath every consumer surface.
Pet ecosystems remain reactive.
They record appointments, transactions, and reminders. They do not observe drift, persistence, velocity, or the subtle emergence of deterioration that precedes a visible symptom by weeks.
Most systems respond only after a visible symptom has already crossed a clinical threshold. By that point, the signal had been present for twenty-eight days.
The ecosystem is evolving toward continuous intelligence.
Wearables, IoT, veterinary EMRs, insurance carriers, shelters, consumer apps and research networks each generate telemetry. None synthesise it longitudinally. The result is a fragmented archive — never a coherent understanding.
Ontology · Foundational
Events are not intelligence. Signals are.
A signal is the synthesis of behaviour over time. It carries persistence, velocity, confidence and cohort context. It is what an event becomes once the system understands the subject.
The biological intelligence layer.
Companion Health is not an app. It is the layer beneath the application — the ingestion, normalisation, synthesis and replay infrastructure that turns raw telemetry into continuous longitudinal intelligence.
- Ingestion · multi-source telemetry, vet records, behavioural input
- Normalisation · canonical schema across species, sources, devices
- Synthesis · longitudinal modelling against the subject's baseline
- Replay · time-scrubbable history of signal emergence
- Confidence · transparent inference weighting on every output
- Intervention support · structured windows surfaced to consumers
The interoperability layer.
Fetch API is the public expression of the ontology. Where Companion is the intelligence, Fetch is the interface — endpoints, signal objects, webhook events, replay queries, confidence-aware responses.
GET /v1/subjects/PET-00472/signals
Authorization: Bearer ••••
200 OK
{
"subject": "PET-00472",
"signals": [
{
"kind": "hydration_drift",
"deviation": 0.49,
"persistence_days": 21,
"velocity": "↑",
"confidence": 0.87,
"window": "14d"
}
],
"cohort": "feline.senior.dsh.n=347",
"as_of": "2026-05-25T11:23:08Z"
}Behavioural baselines, persistence, confidence.
The engine computes deviations against a per-subject baseline, weights them by persistence and velocity, then reconciles against cohort drift. Every output carries a confidence value the consumer can act upon.
System · Nine layers
Production-grade infrastructure.
- L01Data SourcesWearables, IoT, vet EMR, intake forms, behavioural input.
- L02IngestionStreaming endpoints, batch loaders, partner connectors.
- L03ValidationSchema enforcement, identity reconciliation, quality scoring.
- L04Signal EngineDeviation, persistence, velocity, baseline reconciliation.
- L05ConfidenceProbabilistic weighting on every inference.
- L06Risk SynthesisComposite risk surfaces over time.
- L07ReplayTime-indexed history of signal emergence per subject.
- L08Fetch APIPublic surface — signals, webhooks, replay, cohorts.
- L09Ecosystem ConsumersVets, insurers, shelters, IoT, research, pharma.
One intelligence layer. Many consumers.
The same biological signal is interpreted differently depending on the consumer. Insurance is one downstream interpreter — not the identity of the platform.
Continuous care, never alarmist.
Caregivers receive measured context — behavioural drift, hydration trends, activity persistence, routine changes — surfaced calmly, with confidence and an actionable window.
Forensic, longitudinal, decision-grade.
Clinicians receive replayable history, prioritised signals, confidence intervals and intervention windows. Decision-making remains entirely human; the system provides operational awareness.
One downstream consumer.
Carriers use the intelligence layer for early-intervention support, cohort risk analysis, underwriting context and claims mitigation. The insurer is interpreting the same signal the veterinarian sees — through a different lens.
The most under-served operational surface.
Most shelters operate without longitudinal behavioural intelligence. Companion enables adoption readiness, deterioration monitoring, recovery analysis and emotional stability tracking — quietly, at the population level.
The future moat is not an app.
It is continuous longitudinal intelligence.
Applications come and go. Telemetry compounds. The platform that owns the longitudinal record — and the ontology that gives it meaning — becomes the biological awareness layer underneath every ecosystem consumer.