Timothy D. Sweetser

Machine Learning Engineer in New York City

• LinkedIn
• Github
• tsweetser@gmail.com

Experience

StubHub, New York, NY

Machine Learning Engineer. April 2024 - Present

• Recommendations and personalization in Python and C#
• Multi-armed bandits for display artwork optimization
• Improve observability, telemetry, and alerting

Warner Brothers Discovery, New York, NY

Machine Learning Engineer. November 2020 - March 2024

• Reponsible for scaling and innovation of recommender systems and personalization for Max and HBO Max
• Scale personalization algorithms to tens of millions of users using Airflow, Docker, and Spark
• Increase user engagement of personalized recommendations tray using EASE model and XGBoost ensemble
• Move from offline to online scoring using custom Tensorflow Serving implementation

FindMine, New York, NY

Machine Learning Engineer. May 2019 - November 2020

• Increase click through rate by 3% using \(\epsilon\)-greedy bandit
• Use deep learning to identify similar styles based on images
• Reduce costs by $50K/year by automatically replacing out of stock items in expert-curated fashion outfits
• Save $60K/year in cloud computing costs by shifting computation from database to Python layer
• Write a Flask app to collect quantitative fashion ratings from subject matter experts

Rent the Runway, New York, NY

Data Scientist, August 2018 - April 2019

• Develop fit algorithm to boost rental wear rate by +0.01
• Increase user engagement by improving personalized recommendations
• Rewrite production recommendation service using asyncio, an asynchronous Python web framework
• Migrate batch model training from laptops to Jenkins

Stitch Fix, San Francisco, CA

Data Scientist. July 2015 - June 2018

• Develop personalized recommender systems for women’s clothing and implement them in production
• Increased revenue by $8+ million per year by quantifying style
• Use contextual bandits to optimize where and when to show different recommendations
• Develop end-to-end data science pipelines in Spark, Python and R using AWS, Docker, and Jenkins

Publications

• (2024) Towards Understanding The Gaps of Offline And Online Evaluation Metrics: Impact of Series vs. Movie Recommendations

Education

• M.S. Statistics, Stanford University
• B.A. Mathematics, Clark University, Worcester, MA

Programming Skills

• Advanced Python (numpy, scipy, pandas, scikit-learn, pytest, tensorflow, keras, Airflow, PySpark)
• Advanced SQL
• Intermediate C#
• Basic scala and Golang
• Proficient with git, docker, AWS (Sagemaker, EMR), and bash