Quang Nguyen (Harry)

Quang Nguyen (Harry)

Data Scientist

Followers of Quang Nguyen (Harry)742 followers
location of Quang Nguyen (Harry)Adelaide, South Australia, Australia

Connect with Quang Nguyen (Harry) to Send Message

Connect

Connect with Quang Nguyen (Harry) to Send Message

Connect
  • Timeline

  • About me

    Data Scientist @ insightfactory.ai

  • Education

    • Funix university

      2020 - 2021
      Nanodegree Data Science 9.3/10

      Studying foundation and all the courses related to data science such as Maths for Data Science, Database, Machine Learning, Data analysis, Deep Learning

    • University of Adelaide

      2022 - 2024
      Bachelor's degree Computer Science 6.5/7
  • Experience

    • PIXTA Vietnam Co. Ltd.

      Jun 2021 - Dec 2021
      Data Scientist

      Active Learning for High Score Image problem :● At Pixtastock website, users tend to label their images by tags as much as possible to keep their photos on the head of the page when someone is trying to search for images but some of the tags do not even relate to the content of the image so that we will have to classify which are high-score tags of each image● However this project is not only applied to this problem but also to the whole system at PIXTA and we will use High Score Image as a case to experiment and implement the active learning method● We can achieve similar model performance by using only 250K images instead of labeling all 1M images, which reduce a huge production cost Show less

    • University of Adelaide

      Aug 2022 - Feb 2025
      Machine Learning Engineer

      ● Research in the field of adversarial learning with a primary focus on enhancing the robustness ofpatch-like triggers in real-world scenarios attacking several object detection models, targeting a diverse set of object detectors such as DETR, DINO, YOLOv5, Faster-RCNN, and SSD. A website has been made for illustration purpose. This paper has been accepted to ACSAC 24’ (Core Rank A)● Quantify Uncertainty Through Bayesian Learning Methods Promoting Diversity Can Defend Against Adversarial Malware using Stein Variational Gradient Descent, MC-Dropout, variational inference and ensemble. This paper has been accepted to ESORICS 2024 (Core Rank A)● Lead the development of a defense strategy against backdoor attacks, featuring training of a LAMA-Fourier inpainting model and the application of explainable methods like GradCAM++ and Guided Backpropagation to pinpoint triggers and sanitize them using the LaMa-Fourier model for inpainting and reconstructing trojan images into clean images. Show less

    • Deloitte

      Jun 2024 - Jul 2024
      Oracle cloud engineer
    • Insightfactory.ai

      Feb 2025 - now
      Data Scientist
  • Licenses & Certifications