San Francisco Bay Area
Technical lead for the LinkedIn marketing solutions team. My mission is to "help the team win." Some ways that this has manifested are: * Prototyping and early implementation of new features on the marketing platform, including streamlined campaign creation, data acquisition, and agentic capabilities. * Maintaining code and service health, especially with respect to integrations with Microsoft services. * Compliance with regulatory requirements (GDPR, DMA) and industry standards (MRC). * Development and oversight of our design review process. * Review of API design, schemas, and other data modeling concerns across the company. * Talent development, especially around mentorship, evaluation criteria, and our promotion process.
Led a 3-engineer team responsible for segmenting users into business demographic segments, including onboarding, normalizing, and productionalizing data about individuals, companies, and IP addresses. Redesigned the user segmentation data pipeline so consuming systems could benefit from new features and data quality improvements. Responsible for monitoring and improving the quality of our classifications. Member of a 3-person ad optimization team that built and evaluated algorithms for improving conversion rates on display campaigns. Our responsibility included building out systems to add more data to our ad serving platform, writing and maintaining an A/B testing framework, designing new algorithms, and conducting experiments to measure their effectiveness. Primary author of our engineering culture guide at http://dev.bizo.com/culture/.
Primary developer for the data collection and classification systems. Redesigned and refactored the existing classification system to add parallelism and enable us to easily modify our classification pipeline. Maintained and operated these systems in production. Created a title classifier to segment users by job function and seniority. Used Mechanical Turk and Crowdflower to crowdsource a ground truth set, then used both Bayesian and heuristic rules to extend the coverage of the classifier to the full set of inputs.