Data Science Experience

TEKsystems / Fedex

Member of IT support area in the firm's Revenue Services group. About ninety percent hands-on.

Lead machine learning initiative to predict freight shipments that require corrections, which may result in customers withholding payment, excessive account aging, and lost firm revenue.
Develop IT support team's machine learning skills and knowledge as these relate to data staging, model training and testing procedures, interpretation of model results, and algorithm differences among other topics.

Freight Shipment Anomaly Detection in Revenue Services (Supervised and Unsupervised, 90-100 Million Records)

Developed supervised machine learning prediction capabilities to identify freight shipments that require post-invoice corrections. Based on preliminary analyses and current assumptions, the project is expected to yield a 30-50% improvement in revenue change that results from freight shipment corrections.
Wrote complete data staging in SQL / Python and created a logistic regression model to prototype solution approach. Redeployed using Spark SQL / PySpark and neural network model on Azure platform for production solution.
Specified modules for identification of misaligned freight class-weight-charges (using cluster analysis) and correction code co-occurrence (using association rules). Delegated these modules to junior staff members and guided their work.
Skills Used: Apache Spark, PySpark, Spark SQL, SQL Server, Python, SQL; Methods Used: Neural Network, Logistic Regression, Cluster Analysis, Association Rules.

Data Science Training Program with Python

Developed and conducted hands-on training program in Python in context of using supervised machine learning algorithms on AIRBNB data to predict property rental price. Data staging utilized KNN for missing value replacement and text classifier for feature engineering. Instruction method oscillates between concepts and programming exercises, and instruction delivery was 100% remote.
Skills Used: Python; Methods Used: k-Nearest Neighbors, Text Mining / NLP, Regression (LASSO, Ridge), Random Forests.