← All plates
Anatomy · 06 · Reference design · in build

Anatomy of a data platform, one row at a time.

A reference lakehouse on Apache Iceberg with synthetic data: orders from an API and customers from MongoDB, followed through six pipeline runs.

    each disc is one run's commit to that layer · open any database to see its rows
    1Sources checkpoint
    2Bronze append-only
    !bronze.rejects
    3Silver dedup + MERGE
    4Gold KPIs 0active orders
    Iceberg catalog

    Need a platform like this? WhatsApp →

    Reference design with synthetic data (100 orders, 50 customers), not client work. Storage and catalog are a configuration profile: S3-compatible storage locally; a profile for Amazon S3 with a Glue or S3 Tables Iceberg catalog is documented, not deployed.