CV이력서

Basics

Name Gio Jung
Label Machine Learning Engineer, Multimodal ML Systems & LLM Evaluation
Email gio6141@gmail.com
Summary M.S. Computer Science (SFSU). I build and rigorously evaluate multimodal ML and LLM systems, from distributed training on HPC clusters to systematic model evaluation.

Education

  • 2024 - 2025
    Master of Science
    San Francisco State University
    Computer Science
    • Thesis: "Evaluation of Vision Language Models using Item Response Theory"
  • 2016 - 2023
    Bachelor of Arts
    San Francisco State University
    Mathematics
    • Minor in Computer Science

Work

  • Jun 2024 - Present
    Research Assistant
    YouDescribe Lab
    Accessibility research lab advancing video audio description (AD) for blind and low-vision users; leading systematic model evaluation to validate VLMs as raters and describers.
    • Leading key frame selection research to improve VLM-based evaluation and description of video AD.
    • Validated 100+ VLMs as raters across image captioning and video AD by building end-to-end evaluation pipelines with Hugging Face and LangChain-based API integrations, supporting 2 ongoing research submissions.
    • Uncovered model-specific strengths and failure patterns across 7 quality rating dimensions by applying Item Response Theory (IRT) to compare VLM and human rater performance.
  • 2025 - Present
    SpaceTimeGPT (NAIRR Pilot)
    Pittsburgh Supercomputing Center
    Project funded by the NAIRR Pilot, scaling a video captioning model (TimeSformer × GPT-2).
    • Prepared and transferred 60K+ VATEX video clips to PSC HPC clusters, preprocessing frames for efficient distributed data loading at scale.
    • Co-trained a 400M+ parameter video captioning model across 16 NVIDIA H100 GPUs (2 nodes) using DeepSpeed Pipeline Parallelism, achieving CIDEr 63 on the VATEX benchmark.
  • Aug 2025 - Nov 2025
    Software Engineering Intern
    Astro Information Security
    Information security startup; built internal email automation tooling.
    • Designed an email classification MCP pipeline to process 1K+ messages with automated rule-based filtering.
  • 2023 - 2025
    Tutor
    TASC, San Francisco State University
  • 2023 - 2025
    Grader / Teaching Assistant
    Department of Mathematics, San Francisco State University

Projects

  • 2025 - Present
    eBay 2025 University ML Competition
    Top 4 finish: sequence labeling for a German marketplace.
    • Ranked Top 4 by iteratively fine-tuning sequence labeling models for a German marketplace, reaching F1 = 0.94 through systematic preprocessing and model selection.
  • 2024 - Present
    ChillMate
    AI mental health chatbot for accessible emotional support.
    • Built an AI mental health chatbot using React, MongoDB, and a RAG-powered conversation engine, while leading a team of 7 as Scrum Master through full project delivery.

Publications

Poster Presentations

  • Nov 2024
    GPT-4o vs. Humans in Rating Image Descriptions
    GMiS CAHSI Research Poster Competition
    Authors: Gio Jung, Ilmi Yoon
  • Dec 2025
    ADx3: A Collaborative Workflow for High-Quality Accessible Audio Description
    NeurIPS 2025 Poster Presentation
    Authors: Lana Do, Shasta Ihorn, Charity Pitcher-Cooper, Juvenal Francisco Barajas, Gio Jung, Xuan Duy Anh Nguyen, Sanjay Mirani, Ilmi Yoon

Awards

  • June 2025
    National Science Foundation (NSF), 3K
    National Science Foundation
  • Jan 2025
    NAIRR Pilot Resource
    NAIRR
  • Nov 2024
    Google CAHSI IRP Grant
    Google
  • 2017
    Jessie F. L. York Scholarship, 1K
    Jessie F. L. York Scholarship

Skills

Programming & Data
Python
SQL
Bash
pandas
NumPy
data curation
batch inference
ML & Multimodal Systems
PyTorch
Hugging Face Transformers
multimodal pipelines
fine-tuning
RAG
APIs
Evaluation & MLOps
Item Response Theory (IRT)
Kendall's tau
BLEU
CIDEr
METEOR
CLIPScore
frame sampling
DeepSpeed
Weights & Biases

Languages

English
Fluent
Korean
Native