Guozhen Zhu

Hi, I’m Guozhen Zhu. I build AI systems that turn raw observations into transferable representations of human state and context, enabling machines to act more helpfully and responsibly in the real world. My work is motivated by human needs like safety, health, and everyday well-being, and grounded in building learning systems that can move from research into real-world use.

I’m currently the Principal Scientist at Origin AI, where I lead research on foundation-model-based perception and multimodal/temporal learning for ambient intelligence. I build general-purpose representations and study how to adapt them efficiently across tasks such as presence understanding, motion and activity, localization, and health-related sensing. My research has translated into products deployed at scale (e.g., 25M+ routers and 100M+ smart devices) in consumer and security products.

Before Origin, I earned my Ph.D and M.S. in Electrical Engineering from the University of Maryland, College Park, advised by Dr. K. J. Ray Liu. Across academia and industry, I’ve focused on bridging modern ML with real-world perception—turning messy, high-dimensional signals into stable, transferable representations and actionable understanding.

Research interests:

• Foundation models for perception: large-scale representation learning for multimodal & temporal data, with strong transfer across tasks and modalities

• Reasoning & interaction on top of perception: LLMs and agents for interaction, planning, tool use, and grounded decision-making using learned representations

• Human-centric ambient intelligence: understanding human state, behavior, and context to enable safety, health, and spatial intelligence in real-world environments

Outside of work, you’ll often find me on the tennis court , playing the piano , baking something semi-sweet , or soaking in nature’s beauty .

Thanks for stopping by! I’m always open to discussions, new ideas, and meaningful collaborations—don’t hesitate to reach out!