Amundsen is an open-source data discovery and catalog tool created at Lyft and named after the polar explorer. Its focus is narrow and practical: help analysts and engineers find the right data fast. It works like a search engine for your data estate, ranking results by usage so the tables people actually rely on surface first.
Search ranked by usage
Amundsen's signature idea is PageRank-style ranking driven by query logs: tables that are queried often, by many people, rank higher. A search for "orders" returns the canonical, heavily-used orders table at the top rather than a forgotten copy. Each result page shows the schema, descriptions, owners, frequent users, and last-updated/usage stats - enough to decide whether to trust a table without asking around.
Architecture
Amundsen stores metadata in a graph (Neo4j or Atlas) with an Elasticsearch index for search, and ingests metadata via the Databuilder framework from sources like Hive, Snowflake, BigQuery, Redshift, dbt, and BI tools. The model centers on tables, columns, dashboards, and people.
Amundsen vs DataHub vs OpenMetadata
All three are open-source catalogs. Amundsen is the most focused - discovery and search done well, simpler to grasp and deploy, but lighter on deep lineage and governance workflows. DataHub offers a richer real-time metadata graph and stronger lineage; OpenMetadata is a newer all-in-one with catalog, lineage, quality, and governance combined. If your primary pain is "people can't find data," Amundsen solves it cleanly; if you need column-level lineage, governance, and quality in one platform, the other two go further.
Where it fits
- Teams whose main problem is discoverability - too many tables, no way to find the right one.
- Organizations wanting a lightweight, self-hosted catalog without heavy governance overhead.