4,274,348
Chilean public contracts and purchase orders
Cross-checked against 134,163 companies, their partners and their history.
ChileCompra · SII
+ Sovereign Awareness
Economic
& territorial
security
Who controls whom, which counterparty exposes you, which asset is at risk and what depends on the outside. Chile, Peru, Argentina and Bolivia, with each country's data already loaded.
01 · ENTITY ENGINE
Who is who, and with whom
Canonical identity for companies, people and agencies, with the ownership and link network that ties them together.
- Mandatory provenance on every data point
- Network of partners, addresses and contracts
- Questions in natural language
The ontology is the core, not one more layer
Everything running above it —resolution, engines, API and interface— speaks the same schema. What one module discovers, the others see.
Data sources
Open dataGovernment registriesSensors and satellitesClient dataSpecialized sources
↓ Continuous ingestion · batch · streaming · API
Layers, from the inside out
01
Ontology
Iron seed
A governed data model that defines the common language.
- 63 object types
- 82 typed relations
- Validation and mandatory provenance on every fact
- Country packs: Chile, Peru, Argentina and Bolivia
02
Resolution & integration
Dynamo
Resolves and enriches every fact against live sources.
- Entity resolution: tax ID first, name second
- Normalization and deduplication
- Data lineage and traceability
- Uncertain merges go to human review
03
Domain engines
Convection
20 specialized engines that compute, analyze and produce intelligence.
- Briefing, threats and risk
- Economy, trade and projection
- Territory, emergencies and signals
- Integrity, legislation and reports
04
Platform & routes
Flow layer
Over 340 API routes exposing the platform's capabilities.
- Per-route authorization
- Modules enabled per organization and term
- API, SDK and webhooks
05
Experience & visualization
Terrain
Country workspaces and an analyst interface.
- Maps and geospatial layers
- Entity Engine and cases
- Executive view and reports
- Questions in natural language
06
Users & access
Shield
People and AI agents who consume intelligence and make decisions.
- Roles, organizations and sub-accounts
- Alerts and notifications
- Audit of every query (Law 21.719)
Intelligence cycle
1 →
Interpret
The language model routes the question; it never calculates.
2 →
Resolve
The domain engine brings real evidence.
3 →
Confidence
Deterministic score: sources, recency, R² and sample.
4 →
Narrate
Writes only from evidence, inventing nothing.
Result
Evidence · Risks · Options · Actions
The idea
behind
Serendyp
Serendyp is not one more dashboard. It is an ontology of the country — companies, people, contracts, sanctions and territory of Chile, Peru, Argentina and Bolivia in a single governed model — where every fact carries its source and every decision is recorded.
01
One ontology of the country
Companies, people, contracts, sanctions and critical infrastructure of four countries in one model, already connected: what one source reveals, the others see.
02
Who stands behind
Validated partner and link networks, crossed with the tax registry and contracts with the state.
03
Truth with its evidence
Every fact states where it comes from, when it was obtained and how confident it is. Real, estimated and simulated data are marked differently, and the AI only answers from cited evidence.
04
Territory, live
Critical assets and their exposure to wildfire, earthquakes, outages and volcanoes; satellite and maritime watch.
05
Sovereign deployment
Installed on your servers with a local language model; nothing can be switched off from outside.
06
Fully audited decisions
Every query, answer and decision is kept in a chained, tamper-evident record: who asked, why, what evidence they saw and what was decided (Law 21.719).