← All projects
JPMorgan Chase · Data Scientist · 2025
Customer Sub-Segmentation
Customer sub-segments for marketing and targeting, built with K-means clustering and supervised learning on secure, enriched datasets.
Problem
Marketing needed finer-grained customer segments to target campaigns more precisely. That meant combining customer data in Snowflake with bureau and credit history into model-ready datasets, while keeping sensitive attributes protected throughout.
Data preparation
- Built PySpark pipelines on AWS EMR to extract and transform customer data from Snowflake into the input datasets needed for clustering.
- Encrypted sensitive customer attributes and securely transferred the processed data to Amazon S3.
- Enriched the datasets with bureau and credit history information via Amazon Redshift, improving feature richness.
Modelling
- Built customer sub-segments using K-means clustering.
- Followed the clustering with supervised learning techniques.
- Worked on every step of the process, from preparing the data through clustering and modelling to delivering the final segments.
Results
Customer sub-segments now used primarily for marketing and customer targeting, backed by an automated pipeline that loads the enriched, segmentation-ready data back into Snowflake.