Engineering: Selected Work
Google Search,
AI Overviews and AI Mode
2021 – present
Google AI Overviews and AI Mode are LLM generated summaries that provide a snapshot of key information about a topic or question. I work on various aspect of response and RAG quality, from measurement to ranking and evaluation.
Google Search, Jobs
2018 – 2021
Google Search allows users to find jobs posted by employers and providers from across the web. Using the jobPosting standard defined on Schema.org, employers of all sizes can surface their listings to the right users.
Stereotropes
2015
With Yannick Assogba, Jim Vallandingham & Bocoup DataVis
An interactive visualization exploring gender tropes from the TV Tropes community database — specifically the differences in words and associations used to describe male vs. female tropes. Shortlisted for the Kantar Information is Beautiful awards.
Measurement Lab Visualizations
2016 – 2017
With Jim Vallandingham, Peter Beshai & Bocoup DataVis
M-Lab is a research partnership between New America's Open Technology Institute, Google Open Source Research, Princeton University's PlanetLab and others, holding the largest public dataset of global internet speeds. The Bocoup DataVis team built a comprehensive data transformation and visualization system to make this critical dataset accessible to policy makers and researchers.
OpenVis Conf was a two-day, single-track conference about the practice of visualizing data on the web, running annually from 2013 to 2017. Speakers covered data processing, storytelling, visual design, and implementation — bringing together a global community of practitioners.
IBM Many Eyes
2008 – 2011
With Fernanda Viegas, Martin Wattenberg, Frank van Ham, Jesse Kriss, Matt McKeon & Stephen Levy
Many Eyes was a public web platform that allowed users to upload data, create visualizations, and discuss their findings. It served as an experimental platform for testing hypotheses about visualization's ability to spur communication and social interaction — and whether that activity could yield new insights into data.