7:50 · Music video
The Iceberg Open Lakehouse
A beat-synced tour from open storage and table metadata to catalogs, engines, governance, and semantics.
Open videoSemantic layers · Lakehouse · Agentic analytics
The Semantic Lakehouse
Open tables settle where your data lives. A semantic layer settles what it means. Put both in place and every dashboard, notebook and agent answers from the same definitions.
Featured music video
From Iceberg tables to shared meaning
See how an open lakehouse creates the governed, portable foundation that semantic layers and AI systems can trust.
Four ideas this site keeps coming back to
Semantic layer
The executable bridge between physical tables and the questions people ask. It holds the metrics, dimensions and relationships every consumer resolves against, so a word like revenue means one thing across the company and to every AI agent reading it.
Semantic layer in the knowledge baseData lakehouse
Data kept in open files and open table formats on storage you own, with the transactions, schema evolution and time travel a warehouse gave you. Apache Iceberg is how most teams get there, and the engine stops being the thing that holds the data hostage.
Apache Iceberg in the knowledge baseAgentic analytics
Agents do not stop at answering. They plan, query and act. What they believe your business means comes from the context you hand them, which is why governed definitions decide whether their answers hold up.
Agentic analytics in the knowledge baseQuery federation
Query data where it already sits instead of copying it somewhere first. One access point across sources, with joins and filters pushed down to each one, so the semantic layer can span systems that were never going to merge.
Query federation in the knowledge base
Questions people ask first
What is a semantic lakehouse?
A semantic lakehouse is a data lakehouse with a governed semantic layer on top of it. The lakehouse keeps data in open files and open table formats such as Apache Iceberg on storage you own. The semantic layer defines what the numbers in those tables mean, as named metrics, dimensions and relationships. Together they let dashboards, notebooks and AI agents answer from the same definitions over the same open data.
What is a semantic layer?
A semantic layer is the executable bridge between physical tables and the questions people ask. It holds metrics with their exact filters and time logic, the dimensions that slice them, and the declared keys that make joins safe. Because the definitions run rather than sit in a wiki, every consumer that asks for revenue gets the same number.
How is a semantic layer different from a data catalog?
A catalog tells you what data exists, who owns it, and who may see it. A semantic layer tells you what the data means and can execute that meaning as a query. Most lakehouse architectures run both: the catalog governs access and lineage over the tables, the semantic layer governs definitions over the catalog.
Why do AI agents need a semantic layer?
Protocols like MCP connect an agent to your data but carry no opinion about meaning, so an agent pointed at raw schemas infers formulas from column names and reports guesses with confidence. Grounded in governed metrics, it queries an endorsed definition and can cite it. Widely cited text-to-SQL evaluations land near 40 percent accuracy against raw schemas and above 80 percent when the model resolves through a governed semantic layer.
What is Apache Ossie?
Apache Ossie is the incubating Apache Software Foundation project formerly known as Open Semantic Interchange. It is a vendor-neutral YAML specification for metrics, dimensions and relationships, so semantic definitions can move between BI, analytics and AI platforms instead of being locked to one dialect. It launched as an industry initiative in 2025, published its specification under Apache 2.0 in January 2026, and entered the Apache Incubator in June 2026 with Snowflake, dbt Labs, Dremio and Salesforce among its backers.
What is headless BI?
Headless BI runs the semantic model as a service rather than as a feature inside one dashboard tool. Definitions live behind an API that speaks SQL, REST, JDBC/ODBC, Arrow Flight SQL or MCP, so every consumer resolves against the same model and the front end becomes replaceable.
Where Dremio fits
Dremio runs the engine, the catalog and the semantic layer in one place, over Iceberg tables in storage you own. Governed definitions sit beside the data they describe, Reflections accelerate them to interactive speed, and the same access rules cover a dashboard query and an agent's MCP call.
Explore Dremio