My Portfolio

Education

Bachelor of Science in Computer Science

Southern University of Science and Technology (SUSTech)

2022 - Present Expected Graduation: 2026

GPA: 3.75/4.0

National University of Singapore (NUS) SoC Workshop

2024.5 - 2024.7 Summer School - Artificial Intelligence of Things

Grade: A

University of California, Berkeley (UCB) BGA

2025.9 - 2025.12 Visiting Student

Pending

Research & Publications

Research Experience

Condition-Aware Diffusion Model with FiLM-based Multi-Channel Generation

Carnegie Mellon University (Remote) · Advisor: Prof. Min Xu· 08/2025–present
- Implemented a diffusion architecture for 4-channel conditional generation.
- Integrated feature-wise linear modulation (FiLM) across network layers for dynamic condition-aware control and improved fidelity.
- Enabled controllable, semantically guided diffusion synthesis, improving adaptability to external conditions and interpretability of generative behavior.
MLLM Knowledge Acquisition from Educational Videos

University of Nottingham Ningbo China · Advisors: Prof. Jiang Liu, Prof. Jianfeng Ren · 01/2025–08/2025
- Designed Knowledge-RAG, a plug-and-play retrieval module for precise grounding and Q&A.
- Proposed metrics (KHD, PKAI) to evaluate knowledge acquisition and hallucination in MLLMs.
- Built the LVideoQA benchmark with temporal annotations linking answers to supporting video segments.
Intelligent Diagnosis Algorithm Combining Temporal Eye-Tracking Medical Knowledge

Southern University of Science and Technology (SUSTech) · Advisor: Prof. Jiang Liu · 09/2024–01/2025
- Analyzed the contribution of eye-movement signals across different time periods and modeled them as medical prior knowledge.
- Used temporal eye-tracking priors to guide attention for diabetic retinopathy grading.
- Improved both accuracy and interpretability of the diagnosis model.
Research on Blood Vessel Segmentation Algorithm of Retinal OCTA Image

Southern University of Science and Technology (SUSTech) · Advisor: Prof. Jiang Liu · 02/2024–06/2024
- Proposed Resolution Restoration Assisted Segmentation (RRAS) to enhance vessel connectivity and segmentation accuracy.
- Introduced Weighted Vascular Dice (WVD) for more expressive evaluation of vascular segmentation performance.

Publications & Manuscripts

PEMUTA: Pedagogically-Enriched Multi-Granular Undergraduate Thesis Assessment. Under revision at IEEE Transactions on Learning Technologies.
Did the MLLM Learn? Rethinking the Professional Knowledge Acquisition of MLLM from Multi-Disciplinary Educational Videos. Manuscript to be submitted.
Intelligent Diagnosis Algorithm For Diabetic Retinopathy Integrating Temporal Eye Movement Medical Knowledge. Manuscript to be submitted.

Projects

Intro to Computer Vision and Computational Photography

Computer Vision · Diffusion · Nerf

Course portfolio showcasing multiple computer vision / computational photography assignments.
Includes experiment write-ups, visual results, and reproducible implementations.
Links: Website

In-Context Learning in Diffusion Models: A Comparative Analysis with Transformers

Project for Designing, Visualizing and Understanding Deep Neural Networks

Research codebase for studying in-context learning behaviors in diffusion-style large models.
Links: Code

Stack Overflow Data Analysis Website

Project for Computer System Design and Application A

A course final project implemented in Java with an emphasis on object-oriented design.
Links: Code

Video Conference System

Project for Networking Systems

Built a video conferencing course project with core communication features.
Links: Code

Piano Project with FPGA

Project for Digital Logic

An interactive piano-related course project focusing on user interaction and audio behavior.
Links: Code

User matching based on eye image Wechat mini program -- "You are in my Eyes"

Project for Multimedia Information Processing

Course project exploring multimedia processing / representation.
Links: Code

Jungle Project

Project for Introduction to Computer Programming (Java)

A programming course project (game / simulation style implementation).
Links: Code

Awards

First Prize, College Students' Academic Science and Technology Works Competition (Guangdong), 2025
University Merit Student Scholarship (First Prize, <1%), 2023–2025
Outstanding Student & Student Leader (<1%), 2023–2025
Second Prize, CUMCM in Guangdong Province, 2023
First Prize, Guangdong Province University Taekwondo Championship, 2023–2024

Skills

Coding & Tools

Python · PyTorch · TensorFlow · Java · Git · Linux · LaTeX

Languages

English (TOEFL 104) · Mandarin (Native)

Weiyi Zhang 张伟祎

Education

Research & Publications

Projects

Awards

Skills