Princeton University

Software Engineering Intern at Facebook

Rockville, Maryland, United States

- Worked to increase user customization in Topic Feeds, a key predictor of retention- Designed, ran and analyzed an experiment to find the best parameters for the customization unit- Evaluated the precision/recall of two models for predicting user interests on survey data- Improved internal tools relating to user interests using PHP

Facebook

Software Engineering Intern

- Worked on a Java project that generates sentences in many languages to express facts- Improved inflected name coverage in Russian from under 30% in non-nominative cases to 90+%

Google

- Worked on a machine learning-based C++ system for extracting information from tables.- Achieved 99.7% accuracy in classifying table cells as header or data. The state-of-the-art was 92% accuracy (from Fang et al.'s 2012 paper "Table Header Detection and Classification"), but on a different dataset that we couldn't use because it's not public. We used the ICDAR 2013 Table Competition dataset.

Lisha Ruan