AI Demand Forecasting Development at Sony
Led the development of Sony's enterprise AI demand forecasting platform on Azure Databricks. The system is now fully deployed and operational across 280+ product lines spanning global markets (Japan, North America, Southeast Asia). Currently in the performance optimization and operations delivery phase — tuning model accuracy, monitoring pipeline health, and driving measurable forecasting outcomes for supply chain and inventory planning.
Problem Statement
Enterprise demand forecasting at scale requires handling diverse product categories, promotional effects, and market dynamics while maintaining prediction accuracy.
My Approach
Developed a specialized forecasting model using PySpark on Azure Databricks, focusing on handling promotional lifts and irregular demand patterns. Implemented feature engineering pipelines and model validation frameworks.
Key Outcomes
- Improved forecast accuracy for promotional periods
- Scalable solution processing millions of SKUs
- Integration with existing supply chain systems
Tech Stack
Tags
Project Info
- Status
- Completed
- Category
- Professional
- Company
- Sony
- Created
- 2 years ago
- Ended
- Present
Links
Related Projects
Open Source Time Series Forecasting System
Production-grade MLOps pipeline using Apache Airflow for demand forecasting with Prophet, SARIMA, and DeepAR models. Built for retail SaaS applications.