Seattle, Washington, United States
I build discovery and recommendation systems that help millions of people find communities, ideas, and inspiration. I've led product for Reddit’s Home Feed and Reddit Answers relevance teams over the past 5 years, owning ML ranking, candidate generation, evaluation frameworks, and multi-surface deployment for 100M+ daily users.
Joined Reddit’s LLM-powered Q&A product drive to several of its highest-priority 2026 initiatives — bringing a strong user-first perspective and feeds-era experimentation rigor to a generative AI surface. • Partnered with engineering on the transition to Agentic LLM architecture with context persistence, multi-tool reasoning, and dynamic decision-making. Led project scope/product specifications, prompt optimization, evaluation, and internal dogfooding. •Expanded Answers into Reddit’s highest-intent discovery surfaces — the Search Results Page and Post Detail Page (the primary SEO user destination). Focused on preview unit quality, click modeling, and query targeting. • Expanded Reddit Answers to international users via machine translation, launching in Spanish, Portuguese, Italian, French, and German. Coordinated across internationalization, GTM, and product teams to maintain quality and geo-relevance while extending a core seeker experience to Reddit’s fastest-growing markets. •Partnered with Ads to design and launch Reddit’s first Shopping Answers experience, addressing the ~26% of organic queries that have shopping intent. Currently in alpha for electronics.
Primary IC PM for Reddit's Home Feed recommendation system, owning the full ML pipeline across 100M+ daily users — from retrieval through multi-stage ranking, value models, and diversification. Averaged 35+ A/B experiments per quarter with consistent, compounding impact. **Candidate Generation & Retrieval** - Defined framework and mixing strategy across retrieval sources like subscriptions, personalized discovery, broader exploration, and geo signals. - Launched Reddit's first Two-Tower neural models (representations for posts, subreddits) for recommendations (+0.3% sessions, +1% consumption, +0.13% DAU, +1.4% subscribes). - Built real-time responsive retrieval sources driving +0.08% DAU, +1.3% upvotes, +0.16% sessions, +0.5% ARPU. - Designed cold-start distribution (candidate retrieval + final stage insertion) for new posts (+1% comments, +0.5% posts/user). **Ecosystem Health & Diversification** - Value model boosts on ranking heads for long-term feed health: small-subreddit boost (+0.1% DAU, +0.35% subscribe rate), fresh post boost (+8% feed post views, +0.35% sessions), low comment post boost (+0.9% comment rate, +1.5% comments/user). - Replaced rigid re-ranking with probabilistic diversification (+0.36% ARPU, +1–2% subscribes and good visits). - Built geo-personalization stack driving +1.4% sessions, +1.8% local consumes, +0.67% time on site internationally.
• Collaborated across the Video team and teams across Reddit to lead a project to improve the sharing functionality on the streaming product, RPAN by identifying, designing, and running experiments on 2 key features to move core metrics on retention rate, engagement rate, and CTR. • Designed and wrote specifications for a new URL tracking system for RPAN share links that would allow employees to identify visits to the site that were through a share link and gain key insights on metrics such as CTR and reach for the first time. • Launched a video creator research project to survey Reddit users and run a gap analysis between Reddit and popular video platforms. Presented findings to the larger video team and the Senior Product Director of Video who used such insights to gauge Reddit’s Q4 priorities for video on Reddit.
•Work as a teaching assistant for INFO 180, Introduction to Data Science. •Introduce a wide, diverse cohort of students to the preliminary concept of data science and data analysis including basic statistic methods and non-technical tools such as Excel, Tableau and BigML. •Assist professor in lectures by facilitating class activities and evaluating projects/presentations •Provide consistent support to students by holding office hours outside of class and personally mentoring a section of 20 students.
•Worked as a key member on the Safeco Insurance MMA (Manage My Account) Digital Team. •Collaborated with product owner, designers, and developers in 2 week, agile sprints to create a documents hub for customers on the Safeco website using frontend technologies such as Typescript and Angular.