Trust & safety - US Data Security (USDS)

• Implemented machine learning and network models to detect harmful actors on social media, including those participating in coordinated platform manipulation, amplifying disinformation narratives, and sharing extremist and conspiratorial information• Collaborated with investigative analysts to detect state-backed and coordinated information operations, as reported by national news outlets including the New York Times, the Wall Street Journal, and NBC News• Drove data science projects and products from conception to deployment by managing collaboration between machine learning engineers, software developers, product managers, UX designers, internal analysts, and other key stakeholders• Improved internal tooling, processes, and patterns for developing and deploying machine learning models• Mentored junior data scientists and machine learning engineers in their technical skills and knowledge of online trust & safety

• Audited the algorithmic susceptibility of Community Notes—Twitter's crowd-based approach to addressing misleading information on the platform—to various forms of adversarial manipulation• Developed an evaluation suite of ground-truth assessments, descriptive metrics, and synthetic experiments to measure the performance of changes to the Community Notes ranking algorithm• Provided cross-functional, embedded data science support to core teams, such as Home Timeline and Who to Follow

• Researched how people amplify their voices online, particularly within the contexts of hashtag activism, polarization, and misinformation• Worked extensively with social media data using Python (e.g. numpy, matplotlib, scikit-learn, spacy, networkx, etc.), Spark, Postgres, Slurm, Dask, and Bash• Developed open source code for new network science and NLP methods to draw novel insights from social media data• Applied domain expertise and qualitative inquiry during data collection, model development, and analysis to produce robust results• Constructed several automated, reproducible data pipelines for colleagues across multiple research labs to use• Communicated results to interdisciplinary audiences of computer scientists, physicists, political scientists, communication researchers, and many others• Mentored undergraduate and graduate lab members to develop their capabilities as researchers and data scientists

• Monitored coordinated activity on the platform by performing unsupervised clustering on networks with billions of edges• Informed product and policy decisions of the Civic Integrity, Virality Monitoring, and Political Organizations & Society teams on coordinated inauthentic behavior and misinformation

• Developed a novel topic model (CorEx) capable of incorporating domain knowledge through "anchor words"• Extended the existing Python codebase to improve its interface and ease of use• Benchmarked the topic model against unsupervised and semi-supervised variants of LDA• Validated the topic model's performance through new, domain-relevant metrics• Applied the topic model to disaster response articles and clinical health notes to improve downstream document classification

• Wrangled many terabytes of historical social media data using Python and high-performance cluster computing• Applied social network and natural language processing models to research counter-protest deterrence of online activism• Developed an open-source Python package for visualizing word-level differences between text datasets, including sentiment differences

• Taught introductory discrete math and calculus courses as instructor of record

• Conducted research on discrete curvature flows through computational simulations

• Tutored student athletes in precalculus, introductory statistics, and calculus

• Conducted computational research on solving partial differential equations using a novel random walk approximation