By Wayne Gillam / UW ECE News
UW ECE doctoral student Mingfei Chen has received a 2025 Google PhD Fellowship in Machine Perception. This award supports Chen’s pioneering work developing AI systems capable of perceiving and understanding three-dimensional spaces — a capability that could transform robotics, augmented reality, and assistive technologies. Photo by Ryan Hoover / UW ECE
UW ECE doctoral student Mingfei Chen was recently awarded a 2025 Google PhD Fellowship in Machine Perception. This award is one of the most competitive honors for doctoral students in artificial intelligence research today. Google.org awarded this fellowship to Chen to support her work developing AI systems that can sense and comprehend three-dimensional spaces. In late October, Google announced the fellowship recipients in its official blog, The Keyword.
The Google PhD Fellowship Program, now in its 16th year, supports outstanding graduate students who are conducting exceptional and innovative research in computer science and related fields, specifically focusing on candidates who seek to influence the future of technology. The Program provides vital direct financial support for its recipients’ doctoral degree pursuits and connects each Fellow with a dedicated Google Research Mentor, reinforcing the company’s commitment to nurturing the academic community.
Chen’s research: Spatially aware AI
Chen, a third-year doctoral student in the UW NeuroAI Lab, which is directed by her adviser, UW ECE Associate Professor Eli Shlizerman, is developing spatially aware multimodal AI systems that are trustworthy, safe, and human-centered. Her research focuses on enabling AI to perceive and understand three-dimensional spaces — a capability that could transform robotics, augmented reality, and assistive technologies.
“I am very excited about building AI systems that can truly perceive the world — not just through vision and language, but through spatial awareness and more modalities.” — UW ECE doctoral student Mingfei Chen
“Barely a few years into her doctoral research, Mingfei has already built cutting-edge deep learning models that combine sound and vision to create detailed representations of 3D scenes — both real and virtual,” Shlizerman said. “Now, Mingfei is daring to take this even further. She is exploring how machines equipped with deep learning can understand the scenes they perceive. It’s an exciting and bold direction, and the Google PhD Fellowship will empower Mingfei to make it a reality.”
Real-world applications
Google.org is providing over $10 million to support 255 doctoral students like Chen across 35 countries and 12 research domains, committing to a new generation of researchers who understand that accelerating scientific discovery is vital to solving the world’s toughest challenges.
Like humans, spatially aware multimodal AI systems use sensing modalities, such as vision, sound, and motion, to build contextual awareness and an understanding of a three-dimensional space. These AI systems also use other modes, such as language and geometry, to enrich their understanding. This technology could help to make the world more accessible and supportive for people, especially for those with disabilities or limited mobility. Potential applications include:
- Spatial memory assistants: AI-equipped eyeglasses could help a person remember where they placed their keys or track how a room changes over time — using vision, sound, and spatial cues to retrieve useful information from the environment.
- Safety in dynamic environments: Wearable devices could detect approaching vehicles or obstacles outside a person’s field of view and provide directional audio alerts.
- Interactive spatial guidance: AI assistants could help people navigate complex environments by aligning audio cues with visual context. For example, when an assistant says “the object on your left,” the sound could originate from the user’s left side, linking language, vision, and spatial geometry. In public spaces like museums, these assistants could fuse real-time visual recognition with spatial audio to direct visitors toward exhibits and deliver information hands-free, enabling intuitive navigation without relying solely on sight.
- Immersive virtual re-experiencing: Spatially aware multimodal AI systems could recreate real environments for virtual tourism or memory replay. Instead of viewing static images or videos, users could “re-live” dynamic scenes (for example, standing near a landmark like the Eiffel Tower) with spatial audio and 3D geometry that deliver a natural, embodied experience.
Empowering STEM education
Chen is also passionate about contributing to STEM education and entrepreneurship. For the past two years, she has served as lead teaching assistant for UW ECE’s Engineering Innovation and Entrepreneurship (ENGINE) capstone program, mentoring over 100 students and coordinating a team of eight teaching assistants. Through ENGINE, Chen has helped foster interdisciplinary collaboration on real-world engineering projects with leading technology companies. She said she is excited to contribute to similar projects in the future through the University as well as through global collaboration opportunities, such as those the Google PhD Fellowship might provide.
Looking ahead
Chen said that this fellowship gives her freedom to pursue unconventional and challenging research directions without being constrained by short-term trends in the field. She is looking forward to collaborating with Google researchers and continuing to build AI systems that enhance productivity, autonomy, and quality of life.
“I am very excited about building AI systems that can truly perceive the world — not just through vision and language, but through spatial awareness and more modalities,” Chen said. “Humans intuitively combine sight, sound, and context to understand the three-dimensional world around us. Pursuing this direction could unlock a deeper form of perception for AI — crucial for future technologies like smart glasses, spatial assistants, and personal robots.”
For more information about Mingfei Chen and her research, visit the UW NeuroAI Lab website.