Medallion Architecture for Data Lakehouse

When to Use

Use this skill when:

• Designing a new data lakehouse from scratch using Bronze/Silver/Gold layers
• Deciding what belongs in Bronze vs Silver vs Gold
• Writing DDL for any Medallion layer on Iceberg/Parquet (Trino, Spark SQL, dbt)
• Implementing incremental pipeline patterns: watermark, CDC, full-refresh, streaming
• Choosing and implementing a deduplication strategy
• Handling schema evolution, late-arriving data, or data quality gates between layers
• Designing partitioning and sorting strategies per layer
• Structuring Airflow DAGs or dbt projects around the Medallion pattern

Architecture Overview

Medallion organises a data lakehouse into three quality layers. Data flows in one direction: Bronze → Silver → Gold. Each layer applies progressively more transformation and enforces progressively stricter quality rules.