Client: Multi-brand Retailer with 1,500+ stores
Tools: Python, Prophet, PySpark, Snowflake, Tableau
Challenge:
Stock-outs and overstock were frequent due to inaccurate, manual forecasting processes across SKUs and locations.

Solution:

• Centralized data pipeline using PySpark and Snowflake
• Trained forecasting models using Facebook Prophet and ARIMA
• Integrated promotional events, holidays, and weather patterns into feature sets
• Visualized forecasts in Tableau dashboards with actionable alerts

Outcome:
Improved demand forecast accuracy by 38%. Reduced inventory holding costs by 21% across regions.