Enterprise Manufacturing Analytics Platform

Overview

A dbt project running on Databricks that transforms raw ERP data into analytics-ready models. It implements a medallion (Bronze → Silver → Gold → Mart) architecture so cleaned, curated tables can power KPI dashboards, leaderboards, and operational views across multiple manufacturing lines, workforce productivity, pre-production planning, and sales reporting.

Architecture

Data flows through four progressively refined layers, each with a clear responsibility:

• Bronze — raw ingested ERP source tables, landed as-is.
• Silver — cleaned and standardized facts and dimensions (orders, units, sales, scans, employee hours).
• Gold — domain-curated, business-ready models with lead times, production metrics, and remake reasons.
• Mart — analytical views: KPIs, leaderboards, and aggregations consumed by dashboards.

Highlights

• Layered modeling with reusable macros for schema naming and shared transformation logic.
• Data-quality tests and snapshots for slowly changing dimensions.
• Seed-driven reference lookups for reproducible, version-controlled metadata.
• Built on Unity Catalog with managed Iceberg tables for Silver/Gold layers.

My Role

• Took part in the architecture design and deployment of the platform.
• Built a three-tier, AI-powered self-service query application on top of the data layer, letting production teams explore the data they need on their own — without going through the data team.
• Implemented access separation so each role and department can only reach the data it is authorized to see.