Yaoning Yu

I’m Yaoning Yu, a Data Scientist & AI Engineer at Ekohe while doing research under the supervision of Dr. Haohan Wang at UIUC (University of Illinois Urbana-Champaign) Dream Lab. Previously, I earned my Bachelor’s degree in Mathematics from the University of Washington (UW). During my time at UW, I worked with George Manucharyan in topics of Ocean Dynamics, mainly responsible for utilizing ML strategies to solve unknown deep ocean activities. I also worked with Dr. Yuan Tian at the UCLA Security Lab on LLM security.

Projects

• TradingHero: Empowering Traders with AI Stock Analysis - An advanced stock market analysis tool designed to streamline trading processes, reduce information overload, and enhance decision-making for semi-professional retail traders (Honorable Mentions).
• CredAIble - An LLM-agent-powered software that automates file classification and recognition in the underwriting process(Private github repo due to clients’ confidentiality).
• Robustifying LVLMs Against Jailbreak via prompt injection attack – Design inverse perturbation strategies based on CLIP embeddings and attention map analysis to defend large vision-language models (LVLMs) against jailbreak attack via prompt injection and improve alignment robustness.

Research Interest

My research interests lie in Machine Learning, LLM-based Multi-Agent Systems (MAS), Multimodal Large Language Models, and AI Security. Additionally, I’m eager to explore other emerging fields(or any related areas) and apply these advances, especially in embodied AI, to solve real-world problems and deliver tangible and meaningful value to communities.

I am actively seeking research opportunities. Feel free to explore my projects and papers and reach out if you’re interested in collaborating or even about any research ideas! 🚀

About Me

In my spare time, I enjoy reading, exploring the arts and music, and immersing myself in nature. I’m also passionate about various sports and always eager to try new activities. My Chinese name is 于曜宁.