Andrew Ding

MLE @ Meta Ads Ranking, ex TikTok, X

New York, New York, United States

About

Tackles complex problems with an open mind and unconventional solutions; has an eye for good design & untapped user needs.

Experience

  • Machine Learning Engineer II at Meta
    2025 - Present · 1 yr 6 mos

  • Machine Learning Engineer II at X
    2024 - 2025 · 1 yr

  • Software Engineer at TikTok
    2023 - 2024 · 1 yr

  • Machine Learning Intern at TikTok
    2022 - 2022 · Less than a year

    • Optimized ad recommendation and ranking systems on App Monetization team by better predicting user conversion through multi-tasked, wide and deep NN models; 5% AUC gains from ESMM architecture over production in A/B testing • Built and shipped 10 new sparse features for characterizing user external action patterns, increased advertiser value by 5% and user conversion rate by 2.6% in testing; MapReduce, Hive SQL complex queries used for data transformation • Improved transfer learning using KNN clustering to map app store IDs to a unified product ID, led to 9.2% AUC gains • Reduced PySpark job runtime 20x by pioneering a method for feature list normalization after sum-pooling

  • Co-Founder at AUXPAD
    2020 - 2021 · 1 yr

    • Founded notetaking startup using speech-to-text to help slow note-takers with in-person and remote meetings/classes • Launched AUXPAD on Product Hunt; transcribed over 2K minutes of audio for over 100+ verified users (and growing), user-testing shows 30% increase in WPM over traditional notetaking • Spearheaded definition of core product features and use-cases driven by user insights from over 60 interviews with students and professionals from various industries in US and Chinese markets • Recruited and led team on development with Firebase, React, Redux, Typescript; deployed clustered Heroku Express server • Orchestrated automated testing with CircleCI, Cypress, Mocha Chai; merged 100+ PRs, pushed 1.37K commits to production • Engineered splicing algorithm for improving hours-long gRPC audio streams in poor network conditions