Resume

Bacholar of Science, Computer Science | Applied Mathematics

• GPA: 3.92/4.0
• Honor: Dean's Honors List (2021-2025), 8th place in UCLA ACM-ICPC Algorithms Contest (2022)
• Core Courses: Machine Learning, Advanced Deep Learning & Neural Network, Computer Vision, Reinforcement Learning, Natural Language Processing, Data Structure, Algorithms, Linear Algebra, Optimization, Probability and Statistics

Research on Embodied AI in Urban Spaces

• Contributed to asset integration for URBAN-X and MetaUrban, developing and curating high-fidelity 3D digital assets.
• Led large-scale object extraction using GPT-4o, Grounded DINO, and Grounded SAM, ensuring realistic distribution for simulation platforms.
• Integrated MetaUrban into Nvidia's Isaac Lab, successfully transferring and optimizing digital assets.

Research on Generative Model Memorization and Fingerprinting

• Conducted analysis of data memorization in generative models (DDPM, GAN), focusing on the layer-wise distribution of memorization scores using Vision Transformers (ViTs) and Convolutional Neural Networks (CNNs) encoders.
• Developed a novel training-free fingerprinting method for identifying generative models' architecture, leveraging layer-wise memorization score distributions, achieving SOTA performance on model identification accuracy.

Wayne Wu, Honglin He, Chaoyuan Zhang, Jack He, Seth Z. Zhao, Ran Gong, Quanyi Li, Bolei Zhou. (2024). Towards Autonomous Micromobility through Scalable Urban Simulation. CVPR 2025.

Wayne Wu, Honglin He, Jack He, Yiran Wang, Chenda Duan, Zhizheng Liu, Quanyi Li, Bolei Zhou. (2024). MetaUrban: An Embodied AI Simulation Platform for Urban Micromobility. ICLR 2025 (spotlight).

Jack He, Jianxing Zhao, Andrew Bai, Cho-Jui Hsieh. (2024). Layer Choice for Memorization Detection and Fingerprinting for Generative Models. Arxiv.

Text-Guided Image Editing Framework

• Developed an end-to-end image generation and editing framework using PyTorch.
• Introduced a training-free, text-guided semantic object segmentation method utilizing DiffEdit (Diffusion-based semantic image editing), BLIP, and other text-to-image models, achieving state-of-the-art capabilities.

EEG Signal Classification Models Analysis

• Explored various architectures for EEG (Electroencephalography) signal analysis, including CNN, RNN, Transformers, and hybrid models to resolve the complex patterns of brain neural activities.
• Evaluated the impact of various hyperparameters and augmentation, improving classification accuracy by 15%.

Fairness and Factuality of LLM

• Embarked on evaluating and enhancing the performance of Large Language Models in detecting fairness and factuality in textual claims, centered on Phi-2, Mistral-7B, and utilized GPT-3.5-Turbo for evidence generation.

Trip Budget Planning Web App

• Developed a trip budget planning web app using Java, JavaScript, and HTML/CSS, improving data storage efficiency by 30% with Google Cloud integration.
• Focused on backend development and coordinated with the frontend team to create functionalities for efficient storage of user, trip, event, and budget data, doubling capacity and boosting performance by 90%.

Summer Institute: Introduction to Generative AI, UCLA

• Mentored 100+ students, led interactive discussions, and presented advanced AI technologies, resulting in a 30% improvement in learning outcomes.
• Developed and implemented a transformer-based autocomplete system and chatbot; provided support and guidance for over 20 chatbot projects.