Solutions · Warehouse Automation

Generate Your Data Platform Layers.

Amfelt turns source metadata into working data platform layers: schemas, load logic, and historization generated from customizable templates. There is no hand-written pipeline code left to maintain, and when a source changes, running discovery again flags what changed.

Join the waitlist

Template-Based Generation

Customizable templates emit target schemas, load logic, and history for every table, including SCD Type 2.

History Built In

Current and historical views are generated per table, so analytics and audits read the same governed data.

Drift Flagged

Run discovery again and schema drift is detected against the data contract and graded by how much it can break downstream.

Template-Based Generation

Customizable templates emit target schemas, load logic, and history for every table, including SCD Type 2.

History Built In

Current and historical views are generated per table, so analytics and audits read the same governed data.

Drift Flagged

Run discovery again and schema drift is detected against the data contract and graded by how much it can break downstream.

How It Works

From Source Metadata to Governed Layers in Three Steps.

Ingest metadata

Amfelt recommends a table type for every table (master, transactional, or reference data), profiling the data first on selected sources.

Pick a template

Choose or customize a generation template. Amfelt emits the schemas and load logic, with the history mode each table’s type calls for, such as SCD Type 2.

Review and run

Accept the recommendations and commit the configuration, and the layers are deployed and loaded. During a run, the advisor can apply the answers it is most sure of.

See It in Action

Layers Generated for You, Table by Table.

Staging, history, and bridge structures generated per table, with current and historical views exposed for analytics and no merge logic for you to write.

Layering

A layer flow for the Jaffle Shop Warehouse: catalog, marketing and sales in Jaffle Shop on PostgreSQL feed jaffle_stg with 7 staging objects in DuckDB, which feeds jaffle_hst with 21 history objects and then jaffle_bdg with 14 bridge objects.

The sales tables of the jaffle-core job: customers and stores keep SCD Type 2 history, while order_items and orders use Append history, each with its property count and table type.

The DuckDB scd2.sql template in the repository Explorer: template expressions in double braces around an UPDATE that closes changed history rows and an INSERT that adds their new versions.

From the Blog

Every Warehouse Needs History. Almost None Have It.

Historization is one of the oldest concepts in data warehousing. It is also one of the most commonly skipped. Here is why that keeps hurting, and how to fix it.

Read the article

FAQ

Common Questions.

Do I still need dbt or hand-written transformation code?

Not for the foundational layers. Amfelt generates the staging, history, and bridge structures and their load logic from templates. Bespoke business logic downstream keeps working with your existing transformation tools, on top of the generated layers.

Can I customize the generated structures?

Yes. Generation is template-based, inspired by dbt. Historization strategy and the schema prefix are adjustable per job, and SQL templates can be overridden.

What happens when a source schema changes?

Run discovery again and Amfelt compares the source with its contract, grading each change as a patch, minor, or major change and listing it against the source for you to review (the app lists drift for the development environment today). Applying changes automatically under a policy you set is on the roadmap.

Which target data platforms are supported?

PostgreSQL and DuckDB at launch, with the catalog growing demand-first; the Integrations page shows current status.

Build on Generated Data Layers.

Join the waitlist to hear when Amfelt opens to more teams.

Join the waitlist