Connect intelligence to verified outcomes.
VetIOS connects clinical evidence, reasoning, verification, and observed outcomes in an auditable workflow. A versioned inference engine and CIRE publication gates make the reasoning inspectable; later evidence tests the original decision.
VetIOS was founded in Nairobi, Kenya. Veterinary medicine is the first proving ground for a broader biological intelligence infrastructure vision.
Five-month-old incompletely vaccinated puppy with hemorrhagic diarrhea
A synthetic five-month-old mixed-breed puppy with an incomplete vaccination series develops acute lethargy and anorexia, followed within a day by repeated vomiting and hemorrhagic diarrhea. It is moderately dehydrated. The CBC shows leukopenia with neutropenia.
Reported probabilities for the top named conditions. Mass the engine assigns elsewhere is shown as the residual, never redistributed.
- Publication
- REVIEW
- Structural
- REVIEW
- Clinical reliability
- UNVALIDATED
- Differential concentration
- 0.1464
fecal canine parvovirus antigen (ELISA): positive. Confirmed: Canine parvoviral enteritis.
The engine ranked it first before the test. CI regenerates this case and fails if that changes.
reasons: Structural risk → REVIEW
- The pre-test differential is flat: the leading condition holds 13.7% of the reported mass, so CIRE publishes REVIEW.
- Clinical expectation places Acute Hemorrhagic Diarrhea Syndrome in the top three; the engine ranks it 5th.
- Known condition-registry normalization issue: 5 conditions are registered under more than one identifier (Acute Pancreatitis, Gastric Dilatation-Volvulus, Hypoadrenocorticism, Immune-Mediated Thrombocytopenia, Leptospirosis). Normalization is pending.
- Acute hemorrhagic diarrhea syndrome remains a differential. It is classically accompanied by hemoconcentration (PCV often above 60%); anemia is not expected. No PCV is supplied, and low PCV is never used as AHDS evidence.
- A positive fecal antigen is treated by the engine as confirmatory. False positives shortly after modified-live parvovirus vaccination are not modeled.
The engine output for this synthetic case has not yet been reviewed by a veterinarian.
engine clinical_deterministic_multisystem_v1 · semantics categorical_topk_with_other · evaluated 2026-09-26 · regenerated from the engine and compared in CI
How CIRE decidesA decision that can be traced to its outcome.
Typed evidence, explicit reasoning, verification, and outcome-linked evaluation form one auditable workflow.
Cases enter as typed clinical input: species, signalment, presenting signs, history, and structured diagnostic tests.
Typed case evidence
A versioned rules engine ranks conditions and lists the supporting and contradicting evidence for each one.
Versioned reasoning
CIRE checks probability semantics and structure, then publishes NOMINAL, REVIEW, CAUTION, or SUPPRESSED before anything is actionable.
Explicit publication decision
Later observations are linked to the original inference, preserving the prediction, confirmed outcome, and evidence provenance.
Decision-to-outcome lineage
Eligible outcome-grounded cases become replayable evaluations and regression evidence for reviewing a candidate release.
Reusable evaluation evidence
Explore the platform and its verification principles.
Clinical work becomes reusable evidence.
For clinical teams, laboratories, and AI evaluators, the useful record connects a decision to the observations that test it.
Reason through a clinical case
Bring history, observations, and diagnostic findings into a structured case. Inspect hypotheses alongside their supporting evidence, contradictions, and uncertainty.
A recorded inference with its evidence trail.
Connect what happened next
Link confirmatory tests and reviewed outcomes to the original prediction. Preserve the earlier reasoning so later evidence can confirm, revise, or falsify it.
An outcome receipt with provenance and case lineage.
Evaluate the next change
Replay eligible cases against a candidate model, policy, or workflow. Compare results before promotion and retain the evidence behind each release decision.
An executable evaluation and a reviewable release gate.
Follow a reproducible example in ProofLoop. Demonstration fixtures are identified separately from clinical outcome evidence.
A field note inside the control plane.
The video is context, not the product. VetIOS should feel like the clinical data substrate underneath every visible interface: provenance-aware, outcome-linked, and operational before it is theatrical.
API-first, typed, and observable.
Every example below is generated from the canonical partner contract and validated against the runtime request schemas in CI.
One external contract at https://www.vetios.tech/api/v1, authenticated with a Bearer vios_k1_ partner key. These cards, the OpenAPI specification, and the SDK are generated from the same contract.
curl -X POST https://www.vetios.tech/api/v1/inference/differential \
-H "Authorization: Bearer $VETIOS_API_KEY" \
-H "Content-Type: application/json" \
-d '{"species":"canine","breed":"mixed breed","age_years":0.42,"weight_kg":6.2,"sex":"male_intact","region":"Nairobi","presenting_signs":["lethargy","anorexia","vomiting","bloody diarrhea","dehydration"],"history":{"duration_days":2,"progression":"acute","owner_observations":["Lethargy and anorexia began first","Repeated vomiting and hemorrhagic diarrhea followed within a day"]},"preventive_history":{"vaccination_status":"incomplete","deworming_history":"unknown"},"diagnostic_tests":{"cbc":{"leukopenia":"present","neutropenia":"present"}}}'{
"species": "canine",
"breed": "mixed breed",
"age_years": 0.42,
"weight_kg": 6.2,
"sex": "male_intact",
"region": "Nairobi",
"presenting_signs": [
"lethargy",
"anorexia",
"vomiting",
"bloody diarrhea",
"dehydration"
],
"history": {
"duration_days": 2,
"progression": "acute",
"owner_observations": [
"Lethargy and anorexia began first",
"Repeated vomiting and hemorrhagic diarrhea followed within a day"
]
},
"preventive_history": {
"vaccination_status": "incomplete",
"deworming_history": "unknown"
},
"diagnostic_tests": {
"cbc": {
"leukopenia": "present",
"neutropenia": "present"
}
}
}DifferentialDiagnosisResponse {
inference_event_id: string | null
clinical_case_id: string | null
data: object
output_payload: InferenceOutputPayload
global_condition_expansion?: object
latency_ms: number
cire: CireSignals
clinical_actionability: ClinicalActionability
actionability_gate?: object | null
formal_constraint_ledger: object
ranker: string
}curl -X POST https://www.vetios.tech/api/v1/outcomes/contribute \
-H "Authorization: Bearer $VETIOS_API_KEY" \
-H "Content-Type: application/json" \
-d '{"request_id":"33333333-3333-4333-8333-333333333333","inference_event_id":"11111111-1111-4111-8111-111111111111","outcome":{"type":"diagnosis_confirmed","payload":{"label":"Canine parvoviral enteritis","confidence":0.99,"label_type":"lab_confirmed","reviewer_exposure":"unknown","verification_context":{"schema_version":"outcome-verification-context-v1","testing":{"decision":"performed","reason":"Diagnostic verification requested at presentation."},"reference_standard":{"name":"Laboratory report","evidence_refs":["44444444-4444-4444-8444-444444444444"]},"specimen_timing":"unknown"},"lab_result_id":"44444444-4444-4444-8444-444444444444"},"timestamp":"2026-06-06T12:00:00.000Z"}}'{
"request_id": "33333333-3333-4333-8333-333333333333",
"inference_event_id": "11111111-1111-4111-8111-111111111111",
"outcome": {
"type": "diagnosis_confirmed",
"payload": {
"label": "Canine parvoviral enteritis",
"confidence": 0.99,
"label_type": "lab_confirmed",
"reviewer_exposure": "unknown",
"verification_context": {
"schema_version": "outcome-verification-context-v1",
"testing": {
"decision": "performed",
"reason": "Diagnostic verification requested at presentation."
},
"reference_standard": {
"name": "Laboratory report",
"evidence_refs": [
"44444444-4444-4444-8444-444444444444"
]
},
"specimen_timing": "unknown"
},
"lab_result_id": "44444444-4444-4444-8444-444444444444"
},
"timestamp": "2026-06-06T12:00:00.000Z"
}
}OutcomeContributionResponse {
outcome_event_id: string | null
clinical_case_id: string | null
linked_inference_event_id: string | null
calibration_delta: number | null
prediction_correct: boolean | null
deprecated_fields?: object
diagnosis_record_id?: string | null
derived_updates: object
request_id: string
meta?: object
error?: string | null
}curl -X GET https://www.vetios.tech/api/v1/models/card \ -H "Authorization: Bearer $VETIOS_API_KEY"
(no request body)
ModelCardResponse {
tenant_id: string
refreshed_at: string
cards: ModelCard[]
}Evidence that survives a change of model.
Cases, policies, outcomes, and evaluations carry their own provenance across the intelligence workflow.
Structured observations with their source and clinical context
Recorded predictions tied to engine and policy versions
Publication gates, explicit uncertainty, and escalation
Later evidence linked to the decision it evaluates
Case-based checks before a candidate is promoted
Scoped access and consent for evidence reuse
Build a workflow whose outcomes can be verified.
Work with VetIOS on clinical reasoning, laboratory reconciliation, outcome feedback, or AI evaluation. Define the decision, the evidence it needs, and the observation that will verify the result.
DESIGN PARTNERSHIPS AND PILOTS: partnerships@vetios.tech · RESEARCH AND VALIDATION: research@vetios.tech