Medallion Architecture for Executive Dashboards

A mid-size software company had executives making decisions based on dashboards that were slow, unreliable, and often showed conflicting numbers. The root cause was a messy data layer with no clear separation between raw, cleaned, and business-ready data. Queries took 10+ minutes to run, and nobody trusted the numbers.

I implemented a full medallion architecture (bronze, silver, gold) in Snowflake, orchestrated by Apache Airflow and transformed with dbt.

Bronze layer: raw data ingested from production databases, SaaS tools, and event streams. No transformations, just faithful copies with metadata.

Silver layer: cleaned, deduplicated, type-cast data with business logic applied. Automated dbt tests caught quality issues before they reached dashboards.

Gold layer: pre-aggregated, business-ready datasets optimized for Power BI dashboards. Query times dropped from minutes to seconds.

Dashboard query time reduced from 10+ minutes to under 3 seconds

Executive team finally had a single source of truth they trusted

Data freshness improved from daily batch to near real-time (15-minute refresh)

Power BI dashboards served 40+ users across 5 departments

Snowflake, dbt, Apache Airflow, Python, Microsoft Power BI

Solo data engineer. Designed the medallion architecture, built all dbt models across 3 layers, configured Airflow DAGs, and connected Power BI. Delivered in 6 weeks.