Three connected research directions
Our agenda connects model foundations, representation learning, embodied interaction, and deployable systems for rigorous, human-aware multimodal intelligence.
Foundation Models & Reasoning
We build models that connect language, vision, video, audio, and structured signals for grounded understanding and trustworthy generation.
The focus is multimodal representation, event understanding, robust adaptation, uncertainty estimation, and reasoning from evidence.
Embodied Agents & World Models
We develop embodied agents that combine perception, memory, prediction, and planning for navigation, manipulation, and dynamic interaction.
The focus is human-aware navigation, world modeling, robotic manipulation, planning, and reasoning over possible futures.
Mobility AI & Secure Deployment
We design deployable systems for multimodal sensing, real-time analytics, secure edge intelligence, mobility intelligence, and public safety.
The focus is traffic analytics, secure sensing, deployable perception, edge deployment, and decision support with accountable oversight.
Research Opportunities
Current UW students across Seattle, Tacoma, and Bothell can pursue MILab research credit, with instructor approval, through TCSS 499, TCSS 600, TCSS 700, or TCSS 702. Prospective Ph.D. students should apply through the CSS Ph.D. program.
Postdoctoral and research assistant opportunities depend on project fit, preparation, funding, and mentoring capacity. Undergraduates should review UW cross-campus policy; graduate students have no cross-campus registration restrictions.
Applied Work & Media
MILab's applied agenda builds on Prof. Zhi-Qi Cheng's prior collaborations at Carnegie Mellon University, including KAIROS/AIDA, DIVA, PSIAP, Mobility21, CMU MFI, and CMU-AIST navigation. These projects connect multimodal evidence, public-interest applications, and reliable AI systems for complex environments.