• Lead team of 5 Analytics Engineers, delivering 95%+ of key company-level objectives.• Designed and drove implementation of new company-wide data model, reducing core table SLAs by 50% and ETL pipeline costs by over 70%.• Introduced core set of custom Airflow operators that improved Snowflake performance, reduced development friction, and provided industry-leading data quality capabilities.• Deployed the initial version of Looker for Faire's sales organization, resulting in a 6% increase in revenue for managed brands due to improved sales visualizations and recommendations.

• Developed in-house data querying and discovery platform with hundreds of MAU across the company• Lead multiple cross-team initiatives to tackle large problems within the data space• Wrote and maintained machine learning deployment framework which helps reduce time-to-launch for ML models by 60%• Built a number of data pipelines to simplify complex data streams into digestible table formats, allowing engineering and product teams to make data-driven improvements to products across the GitHub platform• Generated and launched a number of machine learning models to streamline internal company workflows and enable popular user-facing features

• Partnered with product managers, data scientists, and engineers across the company to inform product direction with data insights.• Built and maintained automated A/B test analysis platform, enabling quick insights on the multitude of experimental features being evaluated with Uber’s rider app.• Championed data democratization within Rider Experience by distilling complex stream of rider app events into more accessible tables; empowering engineers and product managers to derive insights on user behavior independent of data science bandwidth.• Aggregated rider app events and backend logs to extract leading indicators of problematic user actions, such as trip cancellation or request abort. Engineering utilized predictions to develop features to preempt bad actions in real time.• Promoted once and consistently given more responsibility.

• Collaborated with analysts and stakeholders across departments to develop data-driven engagement marketing strategies.• Developed player churn model which allowed accurate targeting for engagement campaigns, resulting in increased retention and revenue.• Optimized game push notifications by quantifying message impact on player behavior and revenue.• Built and maintained automated re-marketing system, allowing for A/B testing and multiple communication channels (push notifications, email, targeted ads, etc.) to reengage lapsed players.• Developed company’s first dynamic offer system by analyzing player behavior to inform strategy on timing and content of optimal in-game sales.

• Developed new method of binary star system detection that provides results for a fraction of the resources and time required by current techniques.• Wrote Python analysis program using NumPy that determines component masses of binary star systems, impossible for nearly all current techniques, to within 15%. • Parallelized analysis code with OpenCL to handle nearest-neighbor algorithm for thousands of stars and tens of thousands of models, in up to 12 dimensions.

• Created RespawnInto (respawninto.bathompso.com), a web app that recommends future video game purchases based on users'​ current favorites. • Compiled game information, ratings, review text, and comments for over 900 video games from IGN.com, and stored data in MySQL database.• Built recommendation engine using jaccard similarity from user comment activity, and Scikit-Learn TF-IDF analysis of game review text.• Deployed front-end interface with Flask, Bootstrap, jQuery, d3, and AWS.

• Pioneered SupremeGolf’s analytics efforts as a one-time consultant by leveraging web traffic data to provide understanding into customer activity, retention, and demographics across the site.• Compiled customer data from Google Analytics, AdWords, AdSense and Facebook Ads, as well as several months of logs on consumer booking activity.• Combined and analyzed data using Python to quantify previously unknown user behavior on the site.

• Developed automatic pipeline to compile calendar events from a variety of sources (visitor’s bureau, chamber of commerce, etc.) into centrally-controlled city calendar.• Leveraged Python NLTK and Scikit-Learn’s natural language processing algorithms to reduce duplication by detecting similar event titles and descriptions.• Created webapp using Flask to display duplicated and revised events for internal review.

• Wrote and maintained autoscheduler program, which optimally chose nightly targets for observation during the 3rd and 4th iterations of the Sloan Digital Sky Survey.• Connected Python autoscheduler to Postgres database using SQLAlchemy to read information on thousands of possible targets, utilizing data on tens of thousands of past observations.• Developed web API using Flask to easily link autoscheduler to other SDSS webapps and interfaces.