A world-leading research lab in Disaster AI: artificial intelligence for anticipating disaster impacts, understanding evolving conditions, and supporting decisions under uncertainty.
The UrbanResilience.AI Lab at Texas A&M University, directed by Dr. Ali Mostafavi, develops artificial intelligence, including large language models, vision-language models, foundation models, and agentic AI systems, for disaster management and urban resilience. The lab is also the research engine of the Disaster AI Initiative at the Institute for a Disaster Resilient Texas (IDRT), creating AI-powered decision support systems for disaster management. Disasters unfold through connected systems of hazards, infrastructure, and people, and we build AI that reasons across all three: learning transferable representations of hazard and risk, maintaining an auditable picture of conditions as evidence changes, and connecting that intelligence to the decisions communities actually face.
Context graphs, ontologies, and benchmarks that test whether AI reasons correctly over geospatial and infrastructure data, measuring decision value rather than just accuracy.
SYS.01Transferable representations of hazard, risk, resilience, and infrastructure state that transfer across places, hazards, and impact types.
SYS.02An evolving, auditable representation of community conditions that tracks sources, confidence, and contradictions as disaster information changes.
SYS.03Agents that decide which evidence to seek, verify, or escalate next, optimizing for what would change a decision rather than retrieval volume.
SYS.04Modeling how failures propagate through communities and connecting those pathways to real decision points, keeping uncertainty visible.
SYS.05Focused models for our core hazard domains (flood depth and access loss, wildfire risk and spread), which are proving grounds for the broader program.
SYS.06From benchmarks that test whether AI can truly reason about disasters, to foundation-model methods for flood prediction and vision-language systems for damage assessment. A selection of our recent published work.
3,000 rigorously verified questions across eight disaster types. A comprehensive test of whether language models can answer the questions disaster management actually asks.
READ PAPER →Can LLMs orchestrate 26 disaster-response tools into executable multi-step workflows? A benchmark with step-level failure diagnosis across 14 models.
READ PAPER →A concept-to-schema knowledge graph that grounds AI reasoning over disaster data in expert knowledge and connects concepts to the geospatial schemas that answer them.
READ PAPER →Domain-aware coreset selection lets a tabular foundation model match supervised flood-depth accuracy using 0.7% of the training data and transfer to unseen watersheds.
READ PAPER →Zero-shot damage classification from multi-view ground-level imagery, applied to the 2025 Eaton and Palisades fires, accelerating a critical recovery bottleneck.
READ PAPER →A vision for collective human–machine intelligence for augmented resilience and the conceptual blueprint behind our agentic Disaster AI research program.
READ PAPER →We're always interested in exceptional students, postdocs, and collaborators.
Contact Dr. Mostafavi