cv

Click the PDF icon on the right to download my one-page CV.

Basics

Name Hong Tien Nguyen
Label Ph.D. Student, Electrical and Computer Engineering, University of Southern California
Email nguyentienhong1997@gmail.com
Phone 323-787-9669
Url https://hong7cong.github.io
Summary Research interests: human-centered video understanding, temporal dynamics in the context of medical imaging, clinical trials, and seamless dyadic interaction.

Work

  • 2023.09 - Present

    Los Angeles, CA

    Research Assistant
    Signal Analysis and Interpretation Laboratory (SAIL), University of Southern California
    • Develop explainable multimodal models of human temporal dynamics across video and speech, with 10+ publications
    • Partner with 3+ multidisciplinary teams (law enforcement, healthcare, linguistics, psychology) to build AI pipelines

Volunteer

  •
    Activities
    Leadership and Involvement
    • Research mentoring (5+ mentees)
    • CVPR / NeurIPS reviewer (5+ papers)
    • Proposal writing (2) and webpage management (2)
    • Temasek Community Action and Leadership Exchange (2019)
    • USC ECE Research Festival (2024)

Education

  • 2021.09 - 2027.12

    Los Angeles, CA

    Ph.D.
    University of Southern California
    Electrical and Computer Engineering (Advisor: Shrikanth Narayanan)

Awards

Skills

Programming Languages
Python
C/C++
Linux / Shell Scripting
Machine Learning
PyTorch/XLA
Hugging Face
OpenCV
Scikit-learn
Matplotlib
Distributed Computing
HPC
Kubernetes
AWS
GCP
Tools
Coding Agents (Claude Code, Codex)
Git
Conda

Interests

Research
Human-centered Video Understanding
Temporal Dynamics in Medical Imaging
Clinical Trials
Seamless Dyadic Interaction

Projects

  • 2023.09 - Present
    Speech Production on Real-time MRI Video
    Video world models, diffusion synthesis, and multimodal encoders for real-time MRI of the vocal tract.
    • Built Arti-JEPA, a video world model for rtMRI speech production analysis that doubles cross-domain phoneme decoding over pretrained baselines and supports applications to clinical speech (CSL submission)
    • Developed Speech2rtMRI, the first audio-guided diffusion model for real-time MRI synthesis of the vocal tract, with up to 34% of samples misidentified as authentic by expert phoneticians (ICASSP 2025)
    • Engineered a multimodal video encoder for real-time MRI fusing optical flow with 6-channel region-of-interest inputs, reducing phoneme recognition error rate by 5% (ICASSP 2026)
  • 2024.10 - Present
    Temporal Dynamics in Human Speech and Gestures
    Motion-centric video understanding and rigorous evaluation of human behavior models.
    • Developed MOOSE, a motion-centric video encoder for fine-grained human activity understanding (TMLR submission)
    • Improved detection of atypical motor behaviors in autism by 6% accuracy, with enhanced saliency-based explainability
    • Audited evaluation practices in co-speech gesture research, showing learned judges exploit speaker and session shortcuts (0.92 accuracy with no true coupling), and proposed 3 controls for rigorous evaluation (ICASSP 2027 submission)
  • 2024.12 - Present
    Egocentric Officer-Civilian Interaction on Traffic Stops
    Large-scale analysis of body-worn camera footage.
    • Designed a video embedding extraction pipeline and setting selection on 30K+ body-worn videos of traffic stops
    • Analyzed 2K+ annotated videos of multimodal collaborative officer-civilian activity from multiple views
  • 2025.09 - 2026.03
    Multimodal Human Trust Perception
    In partnership with Capital One.
    • Collected and benchmarked the EVALUATE dataset (920 ads), exposing gaps in emotion and trust perception
    • Showed via label-stability analysis that survey design systematically shifts human annotator judgments
  • 2022.09 - 2023.12
    Image Severity Ranking in Ophthalmology
    Expert-centric severity representation learning for medical images.
    • Designed ConPrO, a severity representation learning framework combining contrastive learning and preference optimization, outperforming state-of-the-art methods by 6% in F1 score (CVPRW 2024)
    • Developed SiameN and SEMISE, expert-centric severity ranking models, improving IoU explainability by 6% and ranking performance by 3% in F1 score (ISBI 2025, EMBC 2025, Ophthalmology Science 2025)