Methodology

How Data + AI becomes a decision

A clear, end-to-end view of the lakehouse architecture behind every recommendation.

Step 1Raw Grocery Data
Step 2Bronze Tables
Step 3Silver Cleaned Tables
Step 4Gold Risk Tables
Step 5Databricks Views
Step 6Data + AI Assistant
Step 7Business Actions

Data foundation

Databricks stores all sales, inventory, products, and store data. Raw feeds land in Bronze, get cleaned in Silver, and roll up into Gold tables with consistent business definitions.

Risk logic

Stockout risk is derived from days of supply (current stock / avg daily sales). Waste risk is derived from perishability, days until expiry, and current stock. Recommended actions follow clear, auditable business rules.

Natural-language layer

The user asks in plain English. The assistant classifies the question into one of seven safe intents and only queries approved Databricks views — never raw SQL. Answers come back as summary, chart, table, and recommended action.

Why Databricks

Databricks powers the trusted backend. Gold tables make business logic reusable. Views make dashboards and AI answers consistent. The same lakehouse supports dashboards, storytelling, and natural-language experiences.

Approved Databricks objects

Schema: · Every object below is queried by the app in production.

Business views

    Gold tables

      Validation status

      Reported live from this running deployment.

      Databricks connection

      Checking…

      Approved objects

      0

      Schema