Xiao Liu
Georgia Institute of Technology, GA·Engineering·Early-career
Engineers increasingly rely on data to run power grids, data centers, aircraft and cities. Pure data-driven models can be hard to trust in high-stakes settings, and cities face growing risks from extreme weather and from the spread of artificial intelligence (AI) into their infrastructure.
- Fields
Engineering
Mathematics & Statistics
Computer Science
- Topics
physics-informed machine learning
domain-aware statistical learning
Structure-Exploiting-Preserving
- Grants
- 2 known grants, through Aug 2029
What they work on
- Develop the Structure-Exploiting-Preserving (SEP) framework, which builds data-driven statistical models inside the solution space of the governing physics equations rather than adding physics as an extra constraint.
- Apply physics-embedded statistical models to three engineering cases: environmental processes that threaten utilities during extreme events, thermal behavior of data centers, and structural dynamics relevant to aviation safety.
- Create methods for discovering governing physics equations from data and for dynamic sampling of non-stationary engineering processes, meaning processes whose behavior shifts over time, in collaboration with industry partners.
- Identify vulnerabilities and key performance characteristics of AI-enabled urban infrastructure and mobility systems through the international Resilient-NET network, and set a research agenda for safe interactions between these systems.
- Advance data-driven disaster preparedness, urban digital twins and prediction of compound cascading extreme events by sharing data and platforms through an OPEN-NET platform.
More about the research
This work connects two threads. One is building statistical models that respect the laws of physics and engineering. The other is organizing an international research network on how urban systems can withstand and recover from shocks.
Why it matters
Utilities, data center operators and the aviation industry need models whose predictions they can explain and trust. Physics-aware statistical tools could help utilities stay resilient during climate-intensified extreme events, cut data center energy use, and improve aircraft structural safety. Through the network, city planners and researchers could gain shared data and a research roadmap for keeping AI-enabled infrastructure and communities functioning through floods, storms and other cascading disasters.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Di Shi
New Mexico State University, NM·Engineering
Model multi-scale climate dynamics, including flooding, drought and temperature extremes, and link them to sociodemographic energy maps of three rural communities.
Romulo Meira-Goes
Pennsylvania State University, PA·Engineering
Develop fault detection and diagnostics methods based on supervisory control of discrete event systems, which model a system as a sequence of events, to detect cascading faults and cyber-attacks in data center cooling with rigorous guarantees.
Marin Clark
University of Michigan, MI·Earth & Environmental Science
Develop a scientific framework that characterizes cascading land surface hazards, such as landslides that dam rivers and sediment buildup that promotes flooding, and anticipates how they will change.
The school
Georgia Institute of Technology: 50 labs on PI Pages, $1.5B a year on research, 572 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.