Haoxiang (Steven) Yu, 于浩翔

Empowering the next generation of Multimodal & Contextual AI

I am a Research Scientist at Meta Reality Labs, building scalable AI systems that power the future of AR/VR and smart wearables. My expertise lies in adapting large-scale foundation models for real-world deployment, with a core focus on on-device machine learning, highly-efficient edge inference, and multimodal understanding (including mLLMs and vision-language architectures). I am driven by the challenge of bringing cutting-edge reasoning capabilities to resource-constrained devices with zero compromise on latency or user experience.

My research centers on pervasive/ubiquitous computing & Machine Learning – specifically the collaboration between machine learning models (e.g., federated/collaborative learning), the interaction between models and the environment (e.g., edge AI, context-aware systems), and the synergy between models and humans (e.g., human-in-the-loop learning).

Those works have generated 25+ publications and 5 U.S. patent applications, earning top honors including the IMWUT Distinguished Paper Award, IEEE MASS Best Paper Award, and IEEE PerCom Best Paper Runner-Up. More importantly, my research has demonstrated strong industry relevance, accumulating 300+ citations and influencing patent filings, research publications, and product roadmaps at companies like Amazon, Samsung, and Toyota.

Prior to Meta, I earned my Ph.D. in Electrical and Computer Engineering from the University of Texas at Austin under Dr. Christine Julien, and conducted applied AI research at the Toyota Research Lab.

Outside of my research pursuits, I enjoy cooking , baking , and participating in outdoor activities . I love bringing the world to my kitchen by experimenting with diverse dishes from around the globe.