Joseph Wang

Joseph Wang

Engineering Research Intern

Followers of Joseph Wang484 followers
location of Joseph WangLos Angeles, California, United States

Connect with Joseph Wang to Send Message

Connect

Connect with Joseph Wang to Send Message

Connect
  • Timeline

  • About me

    MS in CS at USC | Ex-Google

  • Education

    • University of Southern California

      2023 - 2025
      Master of Science - MS Computer Science 3.75
    • LeTourneau University

      2019 - 2023
      Bachelor of Science - BS Computer Engineering 3.98

      Activities and Societies: IEEE, ACM, Student Executive Cabinet (LSM President) Software subteam lead for collapsible quadcopter drone for senior design.

  • Experience

    • University of Minnesota

      Jun 2018 - Aug 2018
      Engineering Research Intern

      * Cut data collection time by 60% for live-mice neuro-engineering experiments by prototyping automated Bonsai (C#) scripts.

    • Air Force Civilian Service

      May 2021 - Aug 2021
      Electronics Engineer Intern

      * Pair-programmed with another intern to code 20+ hardware unit tests with 90%+ coverage for digital and analog logic for aircraft simulation circuit cards.* Decoded and debugged 1553 serial communication (MIL-STD-1553) for Boeing B-52 Stratofortress avionics.

    • Google

      May 2022 - Aug 2022
      Software Engineering Intern

      * Developed cast streaming integration test for Chrome on Fuchsia operating system with 14 code changes in open-source Chromium codebase, creating C++ shell browser with JavaScript injection to send video frames and Python GPU pixel test.* Deployed tests to continuous integration (CI) buildbot to run 20+ builds a day.

    • USC Information Sciences Institute

      May 2024 - now
      Student Researcher

      * Trained long-context and state space models (M2-BERT, LongFormer) in multi-node, multi-GPU settings with distributed data parallel to perform dense passage retrieval with PyTorch on legal dataset of 1.84M U.S. caselaw documents and 23.7M passages.* Collected 450k U.S. district court opinions to evaluate 20 masked and autoregressive models across 7 model families (ALBERT, BART, BERT, BLOOM, GPT-2, OPT, RoBERTa) on sentence-pair bias benchmarks.

  • Licenses & Certifications