Loading ParaVida…
Demonstration dataA demonstration portfolio. Every property and every reading was authored to exercise the product. No hotel supplied any of it and no company is described.
Enterprise readiness is not a claim, it is a list — and the useful half of the list is the part that is missing. Nothing below is aspirational, and nothing is described as in progress unless work has actually begun.
Is ParaVida SOC 2 certified?
Not startedNo. There is no SOC 2 Type I or Type II report, no audit in progress, and no auditor engaged.
What it would takeA readiness assessment and a control-implementation period precede any audit. This is a months-long programme, not a document.
Has the platform been penetration tested?
Not startedNo external penetration test has been commissioned or performed.
What it would takeA scoped external test once multi-tenancy and API authentication exist.
Is consumer user data isolated at the database?
In placeRow-level security is enabled on every user-owned table in the consumer product and asserted by an automated suite that attempts cross-user reads and requires them to fail.
Is the API authenticated?
Not startedNo. The read endpoints in this build are unauthenticated and serve synthetic inventory only. They are a demonstration surface, not a production integration.
What it would takeAPI key issuance, scoping, rotation and per-tenant rate limiting.
Is the platform multi-tenant?
Not startedNo. One portfolio is served from source control. A second partner could not be onboarded without building a tenant model.
What it would takeTenant model in Postgres, row-level security on partner tables, and a partner identity separate from the consumer auth system.
Is the intelligence reproducible?
In placeYes. Scoring is a pure function of its inputs — no clock, no randomness, no network, no model. The same portfolio and the same guest always produce the identical number, and a golden fixture suite asserts it byte-for-byte.
What personal data would ParaVida hold?
PartialIn the consumer product: a Passport of stated preferences, saves, rejections and feedback, with export and deletion implemented. The enterprise product in this build holds no guest data at all — it scores archetypes, not people.
What it would takeA partner integration that scored real guests would need a data-processing agreement, a lawful basis, and a decision about whether ParaVida ever receives guest identity. The current design says it should not.
Does ParaVida receive guest identity through an integration?
In placeNo, by design. The compatibility interface takes a weight vector and a property identifier. It has no field for a name, an email, a booking reference or a loyalty number, and adding one would be a design change requiring approval.
How are changes to the intelligence governed?
In placeEvery layer carries its own version — scoring, weights, ontology, climate model, context, engagement — and every result carries the tuple it was produced under. Changing a scoring constant requires a version bump, a regression comparison and founder approval.
Can a model change a ParaVida number?
In placeNo. An AI-proposed reading is retained with its evidence and excluded from the arithmetic until a human confirms it. There is no model in this product and no model credential configured.
What is tested?
In placeScoring units and golden fixtures, divergence, context provenance, destination data validation, database isolation, partner-layer isolation, rendering, and an end-to-end browser suite. The suite blocks merge.
Is there an SLA?
Not startedNo availability commitment, support tier or incident process exists. There is no on-call rotation to commit one against.
What it would takeAn operational model has to exist before an SLA can be offered honestly.
Are commercial terms reviewed?
Not startedNo master agreement, DPA or partner terms have been drafted or reviewed by counsel.
What it would takeLegal review before any pilot involving a partner's real property data.
What is genuinely operational today?
Partial3 of 9 declared capabilities are operational. The other 6 are listed individually with the dependency each one is waiting on, rather than summarised.
Counted from the repository at request time, so the number cannot drift from the truth.
| Suite | Files | What it protects |
|---|---|---|
| e2e | 13 | The browser, end to end, including this enterprise product. |
| experience | 13 | Composed surfaces built on the engines. |
| freshness | 13 | Automated coverage for this area. |
| scoring | 8 | The compatibility model, including golden fixtures asserted byte-for-byte. |
| rendering | 7 | That surfaces render what the engine produced and nothing else. |
| ranking | 6 | Ordering and divergence between different people. |
| feedback | 5 | The feedback state machine and its provenance. |
| unit | 5 | Weights, routing and analytics. |
| destinations | 4 | Curated data validation — incomplete required attributes fail loudly. |
| context | 2 | The trip-context overlay and its weight provenance. |
| discovery | 2 | Automated coverage for this area. |
| dna | 2 | The Environmental DNA geometry. |
| editorial | 2 | Automated coverage for this area. |
| engagement | 2 | Automated coverage for this area. |
| enterprise | 2 | Automated coverage for this area. |
| metrics | 2 | Metric definitions, so they cannot change silently. |
| partner | 2 | The isolation boundary between the enterprise layer and the frozen engines. |
| privacy | 2 | Export, deletion and analytics opt-out. |
| research | 2 | Automated coverage for this area. |
| audit | 1 | Automated coverage for this area. |
| briefing | 1 | Explanation language, including a permanently banned lexicon. |
| cultures | 1 | Automated coverage for this area. |
| districts | 1 | Automated coverage for this area. |
| entities | 1 | The entity adapter and the shared ontology. |
| geo | 1 | Automated coverage for this area. |
| imagery | 1 | Automated coverage for this area. |
| journeys | 1 | Automated coverage for this area. |
| living | 1 | Automated coverage for this area. |
| temporal | 1 | Automated coverage for this area. |
| worlds | 1 | Automated coverage for this area. |
The intelligence is production quality. The platform around it is not, and this page is the list.
The scoring, weighting, evidence and context engines are deterministic, versioned, regression-tested and governed by an approval process. Multi-tenancy, API authentication, verification write-back, SSO, metering and an SLA do not exist — not partially, not behind a flag. A partner engagement that needed any of them would be building them, and that should be priced and scheduled rather than assumed.
Produced underscoring 1 · weights 4 · ontology 1 · climate model 1 · engagement 1 · min coverage 0.5