Problem
- Over- and under-stocking, manual staff scheduling, no real-time sales insight
- Unstructured operational data that made customer segmentation impossible
University & Competitions
Integrated Apparel Store Management System
WinningSolution — a platform that turns inefficient store operations into data-driven automation
Grand Prize, Big Data Specialist Program (4th Cohort) Final Project
Period | 2025.11 ~ 2025.12
Stores with no operational data, run on manual management
My Role
- Built a data generation process from scratch where no operational data existed
- Designed the MySQL ERD with partitions, indexes and foreign keys for large data volumes
- Separated computed and source values and designed integrity checks for reliable data management
- Implemented FastAPI endpoints that automate inventory changes, sales loading and attendance records
- Designed the pipeline from ETL to model (RFM, K-Means, demand forecasting) outputs in the DB, with Jenkins and Docker CI
Result
- An integrated dashboard showing sales trends, sales mix and transactions at a glance for immediate decisions
- Defining the data structure and development standards up front improved the team's development speed and quality
Technical Decisions
Normalized schema with partitions, indexes and foreign keys
Keeps queries fast and data consistent as sales, inventory and attendance pile up
Computed values separated from source values
Source data is preserved and computed results can be verified by integrity checks
Jenkins and Docker CI
Model retraining and deployment run repeatedly without manual steps
