The Opendome core is open source, Apache 2.0, with no license fees. On top of it, value-added services and an enterprise app make infrastructure deployment and administration easier.
Solution = technology + services
The technology: the open-source core that centralizes, describes, and governs your data. Open format, with no lock-in. You can operate it yourself.
The complete data journey: it enters through connectors, is transformed through layers, resolved into identities and corpora, converges in the data lake, is linked in the semantic graph, and is queried under the dome. Each stage leaves the data better described, more traceable, and more governed than the one before.
Connectors extract data from source systems: databases, CRM, ERP, email, files, and voice. Each connector is declared through configuration and versioned with the rest of the project.
From ingestion onward, data is classified by type: structured —records and tables—and unstructured —documents, calls, and emails. Each type follows its own processing path.
Data passes through transformation layers defined for each project: extraction, normalization, cleansing, and enrichment. Each layer leaves the data more complete and consistent than the one before.
At every step, atomic lineage is recorded: the source, date, and transformations of every individual data point. The work is done once, at ingestion, and makes everything built on top traceable.
Business identities emerge from structured data. Entity resolution recognizes that customer 4471 in the CRM, CUST-4471 in the ERP, and c_4471 in a file refer to the same identity, and deduplication merges them into one.
The result is a catalog of unique identities—customers, orders, and contracts—with consolidated attributes and complete lineage.
Content is organized by corpus—calls, reports, and contracts—and each item is processed to infer its metadata: which customer it refers to, its date, and its subject.
This metadata anchor makes every document searchable and linkable: it is the joining point the semantic layer uses to connect it to identities.
Identities and corpora converge in one data lake built on Iceberg and Lance: structured data in versioned tables, and content with its metadata in vector format.
All curated data shares one storage plane, together with its lineage and description. The data lake brings the pieces together; the next layer connects them.
The semantic graph emerges above the data lake: identities connect to one another—Customer places Order—and corpora are anchored to them through their metadata: a call mentions a customer; a report appears in an order.
The ontology emerges from real data: the graph’s concepts and relationships describe what actually happens across the organization’s systems.
A library of access policies governs the entire graph. Every query is compiled with the applicable policy—query′ = compile(query, políticaᵢ)—before it touches the data: rows, columns, and content are filtered within the query itself.
The dome encloses all knowledge: every access is evaluated against a policy and recorded immutably. Automatic redaction extends the same governance to content, so each response delivers exactly what the task justifies.
Knowledge is consumed through queries: a business question enters through API or MCP and returns a response built on the graph, with policies already applied and lineage back to the source data.
People, AI, and systems use the same interface. Any AI, from any provider, operates on knowledge that is already built: it receives resolved context and focuses its capabilities on the task.
The semantic graph in detail: business entities, their meaningful relationships, and the lineage that explains every value from the data that generated it. The ontology emerges from real data and can be queried as knowledge.
The semantic graph arises from the data and relationships in the data lake: business objects and their relationships, now named and meaningful. Customer signs Contract; Contract generates Invoice.
It is not declared by a committee based on what it thinks happens; it emerges from real data.
For every field value and every identity relationship: which system it came from, when it was obtained, and which versions it has had if it changed.
Lineage makes it possible to explain the ontology from the raw data that generated it, auditable all the way to the source.
People, AI, and systems submit queries; every request is compiled with policies before touching any data. Access is decided deterministically at cell level, and everything is stored in an immutable record.
People, AI, and systems access data through a request: who they are, the task they are performing, and what they ask for travel with every request.
Every request is compiled with policies—deterministically, at cell level—and granted or denied. Access control is reproducible: the same request always produces the same decision.
If the data is unstructured, the response arrives redacted: as the complete document, with only the sections justified by the task visible.
An immutable record stores which identity accessed which data and when. It is evidence of what happened, available for any audit.
From where you are today to AI deployed in production. We provide the data infrastructure—semantic graph, response-to-data lineage, and European security—and work alongside you at every stage. AI is deployed with you, using the tools you choose.
click a point to pause
We divide your company’s knowledge universe into galaxies—domains with their own gravity. We also identify important information gaps and propose how to fill them.
The technology foundation, deployed in your cloud or a private cloud, with connectors to your systems and secure inference on open models in Europe.
The semantic graph: resolved identities and business relationships that connect all knowledge, with every data point anchored to what it means.
We define access policies—scoped by identity—and deploy them over the data: the dome. Evidence is audit-ready for GDPR, DORA, and the EU AI Act.
People, systems, and agents access governed knowledge, with every response compiled against your policies and traceable back to the data.
We map what you have, group it into domains, and identify what matters but is missing—with a proposal for how to generate it.
System and data inventory, preliminary identity discovery, and access strategy.
Intelligent clustering into domains with their own gravity: your knowledge galaxies.
We locate important information gaps and propose reliable ways to fill them.
Data completeness, consistency, and reliability as the foundation for all subsequent work.
The foundation deployed wherever you choose, connectors to your systems, and the pipeline that makes data ready to use.
The platform in the customer’s cloud or a private cloud—wherever you choose, under your control.
Inference on open-source models in European infrastructure.
Configuration of connectors to your systems and file systems.
Connectors for proprietary systems not covered by standard integrations.
Definition of the pipeline transformations that make data ready to use.
Resolved identities and business relationships that connect all knowledge, including unstructured corpora.
Identity resolution generation and tuning across your sources.
Generation and fine-tuning of semantic-layer relationships.
AI inference to generate new attributes from existing data.
Access policies are defined and deployed over the data. Evidence is audit-ready for GDPR, DORA, and the EU AI Act.
Identity-scoped access policies, compiled with every query.
Ready-to-use compliance packages—GDPR, DORA, and the EU AI Act.
Audit reports prepared to generate compliance evidence.
Everything shown here can be done with the open-source tools underneath. The console is the interface that makes it automatic, visible, and repeatable—with no loose scripts or knowledge held by one person.
The entire data flow on one screen. We connect your source systems—structured and unstructured—and build the pipeline that normalizes, deduplicates, and distills them.
Each stage is a column with its status and counter: which sources enter, which transformations run, and what lands in the lakehouse.
The destination is an open-format data lake: Iceberg and Parquet for structured data, and Lance for corpus vectors.
Meltano, dbt, and Iceberg run underneath. You could orchestrate them manually; the console connects and monitors them and tells you where every row is.
We resolve identities: the same customer spread across three systems becomes one entity, with a measured confidence level.
We describe business relationships on top of them, and unstructured corpora—reports, calls, and notes—are anchored to the entity they mention.
The console lets you explore, query, and correct the graph. You control the tuning; it is not a black box.
The graph lives in an open format and can be queried externally. The console saves you from writing it by resolving, proposing, and letting you correct it.
Together, we define who can access what, cell by cell and scoped by identity—the same standard for people, systems, and AI agents.
The policy does not filter the response after it is generated: it is compiled with every query and filters the data before anyone reads it.
The corpus is compiled against the requester’s policy: each document gets a version with out-of-scope passages redacted, and each table gets one without denied columns.
Policies are plain text, versioned, and readable. You could write them yourself; the console composes and simulates them and shows you what can and cannot be read.
Every access is written to an append-only record: who, which data, which version, and the exact millisecond.
Lineage ties every figure in the response to the exact cell it came from, including the snapshot revision used.
Denied access is recorded too: evidence that a policy worked matters as much as evidence of the data delivered.
The record is an open table you can query with your own engine. The console reads it for you and produces the audit report.
People, systems, and AI agents access knowledge that is already governed, each within its own scope. If needed, we identify the highest-return areas with you and build agents and insights there using the framework your team chooses. The result: every response is compiled against your policies and written to the immutable record.
Together, we prioritize the highest-return areas and build what is needed there—agents, automations, or copilots—on governed knowledge.
AI in productionWe explore the semantic graph and centralized data to uncover actionable insights that are ready to use.
actionable insightsWhat sustains operations after the journey is complete—including the deployment of new knowledge galaxies—and the commitment almost no one puts in writing.
We support your operation with committed severity levels and response times.
We keep your deployment up to date with security patches applied.
We train your team to operational independence. We do not withhold knowledge.
We review performance, architecture, and security with you regularly.
We deploy new knowledge domains as the business needs them: another area, another system, another corpus—all connected to the graph.
Data, models, and configuration remain portable. You can leave whenever you choose.
Organizations that contact us are looking for a technical conversation about preparing their data infrastructure before deploying AI. If that is where you are, let's talk.
Request a meetingSix European regulations apply to an enterprise AI deployment, and most of them ask for the same things in different words. This manual reduces them to four common questions and explains what each one requires, of whom and from when. It covers the four layers of data sovereignty. A shared reference for executives, business, technology and legal teams.
Descargar el whitepaperBusiness memory already exists: it lives in the data. The only memory that should belong to the agent is procedural.
guardrails vs data security"The agent has guardrails, so the data is protected." That sentence conflates two things operating on different planes. Guardrails watch what the agent says; they don't govern what it accesses. Only structural, reproducible access control protects the data — because it isn't statistical: it's deterministic.
Enterprise AI strategyThere's no enterprise AI strategy without centralizing, relating, describing, and protecting the data. They aren't four good ideas to choose among. They're four conditions, and missing one invalidates the rest.
Ver todos los recursos