|
|
Designed and implemented a channel recommendation model to identify the most relevant YouTube channels for a brand based on campaign keywords and URL-derived context. The approach leveraged shared embedding spaces and novel clustering techniques to account for multimodal channel content, paired with a two-stage ranking system optimized for real-time querying at scale. This system reduced channel selection time by ~90%, reproduced expert decisions with >99% precision, and was patented.
2018-2021Applied Monte Carlo simulation techniques to estimate aggregate view outcomes for planned video lineups, addressing systematic underprediction caused by outlier effects. The approach enabled reliable percentile-based forecasting, was adopted in a patented solution, and improved planning accuracy for internal stakeholders.
2018-2021Developed a UI-driven pipeline to streamline and automate video review workflows.
2018-2021Implemented a service to support CRUD operations, exports, and reversals for a payments table.
2018-2021Worked with audience sentiment data to aggregate and analyze signals at the channel level in support of multiple backend initiatives.
2018-2021I did some side work on a project which attempted to automatically process contracts. For my part, I scraped the FCC EDGAR database to find contracts to be used by human and machine labelers.
2018-2021