• Implemented predictive analytic solutions focused on customer relationship for telecom client with over 4 million subscribers. Project included full-stack development and integration of customer churn/attrition prediction, lifespan, lifetime value, segmentation, and call analysis.

• Developed Hadoop prototype for analysis and storage of call detail records (CDR).

Methodologies:

CRISP-DM, Functional programming, MapReduce

Technologies:

Oracle Data Mining (ODM), Oracle Database, Oracle Communications Data Model (OCDM), Python, Clojure, Neo4J, SAP Business Objects, sed, grep, awk, Hadoop

• Member of team that developed a real-time reusable data API to ease transition to big data and allowing real-time lookup and aggregation of past transactions in company's flagship fraud detection engine.

• Developed and implemented a library to interface with RapidMiner and apply models created in RapidMiner directly in a JVM application.

• Developed banking specific RapidMiner GUI plugins facilitating training of models.

• Implemented CRISP-DM data mining process for company predictive modeling projects.

Accomplishments:

• Saved company over $10,000 per installation by eliminating vendor fees.

• Increased success rate of predictive modeling projects by better aligning business objectives with data mining goals.

• Real-time lookup in fraud engine allows more complex data relationships and transformations to be used in predictive models, thus easing modeling constraints and increasing precision.

Methodologies:

CRISP-DM, Object oriented design patterns

Technologies:

Java, R, Python, Rapid Miner, Voldemort, MongoDB, DynamoDB, CouchDB, EC2, S3, SVN, REST, JSON