Yiyi Wang Email & Phone Number

Kirkland, Wa, Us

San Francisco, California, Us

- Revitalized Nextdoor's search product, significantly improving its year-over-year performance.- Led and developed a comprehensive opportunity sizing framework, enabling data-driven decision at the company level.- Established key metrics, data pipelines, and business and diagnosing dashboard from ground up.- Designed and analyzed all search experiments, elevated the experiment knowledge level across product and engineering teams via weekly meetings, presentations.

Reflect, adjust, and regain momentum

San Francisco, Ca, Us

- Tech lead in relevance platform DS initiatives. Build vision and roadmap and align with product managers and engineer partners. - Initiated signal prioritization strategy for candidate generation (CG). Studied core user signal coverage and intensity by double machine learning (DML) for all Twitter customers. The newly identified signals and new ways of CG is A/B tested and increased 500k mDAU. - Launched offline evaluation benchmark which predicts online experiment performance by offline metrics. This was employed by multiple teams to fail fast and save development time.- Led an opportunity sizing for a company-wide initiative. Provided guidelines of setting goals for the Personalization initiative for the year 2022 to 2024.

Seattle, Wa, Us

Amazon Alexa AI:- End-to-end owner of the first-ever automated AI-product in the dialogue evaluation domain. The model has been A/B tested to effectively reduce human annotation effort by 10%, and been deployed to 7 different language-speaking markets worldwide. - Manage the project by getting alignment with stakeholders, seeking resources to help removing blockers, decomposing the project into smaller tasks, and building milestones and launch plans.- Led the initiative of automating goal evaluation workflow in calculating the entitlement and developing automated AI solutions. The results showed a reduction of human annotation time by 30% with 95% precision. - Led a summer intern to develop a novel approach to evaluate the label quality by leveraging meta learning with noisy labels. The paper was accepted by ACL 2022 ECNLP Workshop.Amazon Payments:- Proposed and prototyped a hurdle modeling approach to solve rare events prediction. Offline experiments demonstrated significant lift (as high as 25%) in AUC for contents with extremely low conversions.Amazon Appstore:- Propose a new marketing budget allocation in absence of A/B testing. Increased annualized Appstore profit level by 2x. - Owned end-to-end dynamic pricing system for Amazon Coins. Conducted experiments to prove the ML system generated 7% more coins sold annually compared to manual decisions. Both papers were published at Amazon ML conference.

Boston, Ma, Us

- Led in making a business case to choose a software vendor for a company-wide anti-fraud solution.- Employed advanced machine learning and visualization techniques to improve customer retention and fraud detection.

New York, Ny, Us

Doctor Of Philosophy (Ph.D.), Statistics

Bachelor Of Science (B.S.), Applied Mathematics