Xidong Wu

Xidong Wu

Research Assistant

Followers of Xidong Wu3000 followers
location of Xidong WuUnited States

Connect with Xidong Wu to Send Message

Connect

Connect with Xidong Wu to Send Message

Connect
  • Timeline

  • About me

    MLE @ Google | RS @ ByteDance | PhD@ U of Pittsburgh | MS @ UC Berkeley

  • Education

    • Politecnico di Milano

      2018 - 2018
      Exchange Program Computer Science

      · Fully Scholarship· Courses:Data structure; Fundamentals of Information System

    • University of California, Berkeley

      2019 - 2020
      Master's Degree Systems Engineering
    • University of Pittsburgh

      2020 - 2024
      Doctor of Philosophy - PhD Computer Engineering 4
    • Beihang University

      2014 - 2018
      Bachelor's Degree 3.86

      - President's Award; - National Scholarship in 2015 & 2017

  • Experience

    • The University of British Columbia

      Jul 2016 - Aug 2016
      Research Assistant
    • University of Cambridge

      Jul 2017 - Jul 2017
      Summer Student
    • Bosch Center for Artificial Intelligence (BCAI)

      Jun 2023 - Sept 2023
      Machine learning research intern

      1. Studied the application of foundation models in federated learning.2. Fine-tuned the CLIP models (RestNet50 and ViT-Base/32) with prompt tuning and linear probing. Compared the ability of foundation models to mitigate client drift with that of other model structures (ResNet, MobileNet).3. Leveraged the zero-shot performance of foundation models to guide the training of lightweight models with knowledge distillation. It largely improves lightweight model performance under severe heterogeneous data settings and benefits inference efficiency.4. Achievement: (1) one paper accepted by NeurIPS2023@FL workshop. (2) One patent. Show less

    • UPMC Hillman Cancer Center

      Sept 2023 - Mar 2024
      Machine Learning Researcher

      1. Used GPT-4 to offer a end-to-end Kowledge graph construction and reasoning in drug discovery.2. We introduced a context-aware relational graph that effectively embodies the hierarchical structure of Molecular Regulatory Pathways (MRPs) and resolves semantic inconsistencies by embedding context directly within the relational edges.3. We used Retrieval-Augmented Generation (RAG) based on constructed MRPs to imrpvoe GPT-4’s reasoning ability. 4. We further developed a novel G-Eval scheme, leveraging GPT-4 for annotation-free performance evaluation on graph construction that demonstrated agreement with evaluations on the benchmark dataset. Show less

    • ByteDance

      Apr 2024 - Nov 2024
      Research Scientist

      Recommended and Search System in tiktok shop

    • Google

      Dec 2024 - now
      Machine Learning Engineer
  • Licenses & Certifications

  • Honors & Awards

    • Awarded to Xidong Wu
      Outstanding graduates in Beijing Beijing Education Bureau Jul 2018
    • Awarded to Xidong Wu
      Shenyuan Prize Beihang University Dec 2017 The highest honor for undergraduate students of BUAA (10/14,000)
    • Awarded to Xidong Wu
      Merit Student in Beijing(Top 1%) Beijing Education Bureau October 1, 2017
    • Awarded to Xidong Wu
      Chinese National Scholarship Ministry of Education of the People's Republic of China (MOE) October 1, 2017
    • Awarded to Xidong Wu
      Second Prize in the Beijing Building Structure Contest The Education Bureau of Bejing Jun 2016
    • Awarded to Xidong Wu
      Chinese National Scholarship Ministry of Education of the People's Republic of China (MOE) October 1, 2015
    • Awarded to Xidong Wu
      Third Prize in the Beijing Physics Competition Beijing Education Bureau Oct 2015