A complete, end‑to‑end analytics engineering project demonstrating professional data modeling, testing, and warehouse design using dbt, DuckDB, and both relational and NoSQL data sources.

This project builds a fully functional analytics warehouse that integrates relational customer/order data with a semi‑structured NoSQL event stream. It demonstrates how modern analytics engineers design reproducible pipelines, model raw data into clean dimensional structures, and validate data quality using dbt’s testing framework.

The warehouse includes:

• A dedicated staging layer with explicit typing and naming conventions
• A Kimball‑style star schema with conformed dimensions
• A fact table that flattens nested JSON metadata
• Automated schema tests for primary keys, foreign keys, and not‑null constraints
• A cost‑free local warehouse powered by DuckDB

This project is intentionally lightweight but architecturally complete — ideal for interviews, portfolio review, or technical discussion.

Raw Sources (SQL + NoSQL)
    |
    v
Ingestion / Extraction Layer
    |
    v
dbt Staging Models (cleaning, typing, JSON normalization)
    |
    v
Star Schema (dimensions + facts)
    |
    v
Analytics-Ready DuckDB Warehouse

• dbt
• DuckDB
• Python
• SQL + JSON
• Kimball dimensional modeling

• Customers
• Orders
• Order items
• Products

A semi‑structured event stream containing logins, support tickets, device changes, location/IP metadata, and nested metadata objects.
The project demonstrates how to normalize, flatten, and model this JSON into an analytics‑ready fact table.

models/
  staging/
    stg_customers.sql
    stg_orders.sql
    stg_order_items.sql
    stg_products.sql
    stg_nosql_events.sql

  marts/
    core/
      dim_customers.sql
      dim_products.sql
      dim_event_types.sql
    facts/
      fct_orders.sql
      fct_order_items.sql
      fct_events.sql

extract/
  Python ingestion utilities (optional)

nosql/
  events.json (NoSQL event source)

sql/
  SQL scripts used during early development

snapshots/
tests/
.gitignore
README.md
dbt_project.yml
requirements.txt
architecture.txt

• stg_customers
• stg_orders
• stg_order_items
• stg_products
• stg_nosql_events — parses JSON, normalizes nested metadata, exposes flattened fields

• dim_customers
• dim_products
• dim_event_types

• fct_orders
• fct_order_items
• fct_events — event‑level grain with flattened metadata such as IP, location, ticket details, and device transitions

All dimensions and facts include dbt schema tests for:

• Primary key uniqueness
• Not‑null constraints
• Foreign‑key relationships
• Event type validation

python -m venv .venv-analytics
.venv-analytics\Scripts\activate

pip install -r requirements.txt

dbt build

This will:

• run all staging models
• build all dimensions and facts
• execute schema tests
• validate the warehouse end‑to‑end

The project uses DuckDB as the analytical engine.
All tables and views are created inside the DuckDB file defined in profiles.yml.

This keeps the entire project:

• cost‑free
• portable
• reproducible
• easy to run on any machine

• Add a date dimension
• Add incremental models for larger datasets
• Add example analytical queries
• Add a metrics layer or BI dashboard
• Add automated ingestion for event streams

This project is designed as a portfolio‑ready demonstration of analytics engineering fundamentals.
It highlights your ability to:

• integrate SQL + NoSQL sources
• design a clean staging layer
• build a star schema
• flatten nested metadata
• enforce data quality
• use dbt professionally
• architect a reproducible warehouse

It is intentionally scoped to be small, clear, and technically impressive — ideal for interviews and technical conversations.