Dashboard: CRM and analytical system for an online store

The online store operates across several services simultaneously: orders, advertising, finance, logistics, analytics. While this data is scattered, the manager sees the business with a delay. I created an internal CRM and dashboard for a fragrance store (two brands, over 82,000 orders in history) to consolidate everything into one framework and brought the system to a working state. Now, the picture doesn't need to be pieced together manually from different tables.

What's inside:
- 36 tables, 186 API endpoints, server interface, visualization through Metabase.
- 8 integrations: OneBox CRM, Nova Poshta, Facebook and TikTok advertising via Airbyte, NovaPay registers via Gmail, NBU exchange rates, OpenAI.
- Automatic matching of about 15,000 payments with orders.
- 25 automatic data integrity checks and role-based access at three levels.
- Determination of customer gender using a directory of 10,514 names, with less than 0.5% unknown across the entire database.

AI works where content needs to be read: one agent analyzes advertising through SQL views and highlights issues in the funnel, while another answers questions about sales and finance. Calculations, turnover, and margin formulas are kept in code to ensure the numbers are predictable.

#Python #FastAPI #PostgreSQL #CRM #Automation #AI #Dashboard #DataAnalysis #SQLAlchemy #Alembic #Metabase #Docker #Railway