Work.
Gatorade
Swift Agency, Portland
Led social analysis around Gatorade’s PG4 collaboration with Nike (Paul George’s signature shoe). Communicated learnings to clients and internal creative and strategy teams to inform the timing, visual language, and copy of colorway announcements. Set new channel highs for engagement on Instagram.
Led monthly @gatorade owned social content reports, making recommendations based in data to drive increases in engagement.
IBM
Swift Agency, Portland
Data visualization and story creation for IBM Data’s social accounts. Scraped, cleaned, and visualized data for an audience of data scientists. Advised designers on charting best practices.
Through social listening and owned content analysis, surfaced insights to inform IBM’s master brand strategy.
Google Pixel
Swift Agency, Portland
Led monthly Google Pixel and Pixelbook social listening reports, providing internal and client marketing teams with actionable recommendations based on consumer perception.
Informed launch strategy for Pixel 3 and 3a product lines, leveraging insights gained from social listening and sentiment analysis around Google and competitors.
Performed ad-hoc analysis for campaigns, competitor product launches, and crisis reporting.
Google Social Lab
Swift Agency, Portland
Spearheaded creative project aimed at making Google’s Twitter more humorous and socially savvy. Wrote various posts for @Google, including its most-shared, most engaged tweet to date.
Delivered monthly Google Assistant social listening reports, identifying trends in the digital assistant landscape as a whole.
Performed various ad-hoc analyses, including (but not limited to): custom sentiment analyses, celebrity partner selection, and models for image analysis, using AutoML in its alpha stages.
MUS Research and Economic Development Initiative
University of Montana
As part of a senior capstone project, collaborated with two fellow Montana mathematics students to search for a biomarker to predict post-traumatic epilepsy, using samples from massive sets of raw electroencephalogram data collected from 48 patients. Classified control and test groups using a random forest algorithm, fine-tuning based on sensitivity/specificity to reach 79% prediction accuracy.
Report (pg. 14-18)