☕️YiFan Cai
(he/him)
Graduate student
I am a graduate student in Systems Engineering at the University of Pennsylvania, with a B.S. in Computer Science from ShanghaiTech University.
I have research experience in diffusion models, computer vision, robotic manipulation, and molecular drug design. My current interests focus on world models, computer vision, and artificial intelligence.
Education
ShanghaiTech University
Bachelor of Science in Computer Science
Key focus: 2D Diffusion
University of Wisconsin-Madison
Exchange Student
Key focus: Intro to Cryptography, Software Security, HCI
University of Pennsylvania
Master of Science in Systems Engineering
Key Focus: World Models, Video Diffusion
Leadership & Service
Chinese Students and Scholars Association at Penn (CSSAP) – Vice President, Public Relations Department
Participated in student community service and event organization,
supporting the well-being and integration of Chinese students.
Volunteer Association (ShanghaiTech University) – Head of Activities Department
Led and coordinated multi-year volunteer programs at ShanghaiTech University,
managing large-scale public service teams and community outreach initiatives.
Shangdao College (ShanghaiTech University) – Mentorship Group Leader
Coordinated college-level academic and mentoring activities, supporting
student well-being and serving as a communication bridge between faculty
and students.
Student Union (ShanghaiTech University) – Member, Life & Welfare Department
Served as a student representative in campus life and welfare affairs,
supporting large-scale events and student–administration communication.
Language
Python
C/C++
Matlab
R
JAVA
AI / ML
Diffusion Models
Computer Vision
Reinforcement Learning
World Models
Probabilistic Models
100%
Chinese
Native
90%
English
Fluent📚 My Research
My research explores diffusion-based generative models as world models for perception, prediction, and decision-making. I focus on three closely related directions:
- Diffusion-based image restoration, including controllable inpainting and completion.
- Generative diffusion models for small-molecule drug design, guided by structured and non-differentiable objectives.
- World models for robotic manipulation, learning latent dynamics that support planning and control.
Across these settings, my broader goal is to understand how generative models can bridge perception and action by providing rich, structured representations of the environment.
Recent Publications
(2025).
Structure-Based Drug Design via Diffusion Models Guided by Non-Differentiable Metrics.
Undergraduate Thesis.
(2024).
Core Technologies in Recommender Systems: Investigating and Analyzing Standard Implementations.
CMLAI 2024.
(2024).
Guidance with Spherical Gaussian Constraint for Conditional Diffusion.
ICML 2024.
Industry Experience And Research
Alibaba Group - Tmall Campus
Development and optimization of generative image stylization models based on GAN and diffusion architectures for portrait-to-anime image generation.
•
1 min read
GRASP Lab, University of Pennsylvania
Research on 3D scene understanding, temporal modeling, and world models for robotic manipulation.
•
1 min read
YesAI Lab, ShanghaiTech University
Research on diffusion models for image generation, computer vision, and molecular generation.
•
1 min read
Selected Projects
PERoKF: Physics-Enhanced Super-Resolution of Kolmogorov Flow
Single-frame fluid super-resolution with physics-consistency loss and physics-derived features
•
1 min read
FEM for PDE-Constrained Optimization in Heat Conduction
Applying finite element methods and convex optimization to sequential PDE-constrained problems in heat conduction (SAND-style formulation).
•
1 min read
HeartAI: Heart Rate Game AI
An AI agent for Heart Rate Game combining reinforcement learning and physiological control strategies.
•
1 min read
Object Detection in Counter-Strike 2
Applying and comparing object detection algorithms in Counter-Strike 2 for reliable player/weapon detection under complex in-game scenarios.
•
1 min read
Building an Intelligent Pac-Man Agent
Incrementally upgrading a Pac-Man agent from rule-based planning to probabilistic reasoning and reinforcement learning, forming a complete AI decision-making pipeline.
•
1 min read