Ali Mostafavi

Ali Mostafavi (Persian: علی مصطفوی) is an Iranian‑American civil engineer, scholar, and technology entrepreneur. He holds the Zachry Endowed Professorship in the Zachry Department of Civil and Environmental Engineering at Texas A&M University and directs the UrbanResilience.AI Lab. Mostafavi is widely recognized for pioneering the use of artificial intelligence, machine learning, and complex‑systems theory to improve disaster preparedness, mitigation, and recovery across urban infrastructure systems.

Mostafavi was born in Tehran and raised in Iran before moving to the United States for graduate studies. Details of his early education have not been publicly disclosed.

After joining Texas A&M University in 2016, Mostafavi quickly advanced from assistant professor to a named professorship and, in 2023, became Resilience Fellow of the 4TU Resilience Engineering Center in the Netherlands. He leads multi‑disciplinary projects with federal, state, and industry partners and has secured more than $38 million USD in competitive research funding since arriving at Texas A&M University.

• Professor, Zachry Department of Civil and Environmental Engineering, Texas A&M University (2024)
• Courtesy appointment, Professor, Texas A&M University Department of Computer Science and Engineering (2025)
• Zachry Endowed Career Development Professorship, Texas A&M University College of Engineering (2021)

• ASCE Walter L. Huber Civil Engineering Research Prize (2025)^[1]
• ASCE ASCE Daniel W. Halpin Award (2023)
• Dean of Engineering Excellence Award–Associate Professor Level (2023)
• Listed in the Stanford–Elsevier Top 2% Scientists database for Civil Engineering and AI (2021–2023)
• College of Engineering Excellence Faculty Award (2021)
• Truman Jones Award for Excellence in Graduate Teaching (2021)
• Zachry Endowed Career Development Professorship, College of Engineering (2021)
• Dean of Engineering Excellence Award–Assistant Professor Level (2020)
• National Science Foundation CAREER Award (2019–2024)
• Engineering Genesis Award (Texas A&M Engineering Experiment Station) (2018)^[2]
• Amazon Web Services Machine Learning Award (2017)
• Early-Career Research Fellowship, National Academies of Sciences, Engineering, and Medicine, Gulf Research Program, (2017)
• Multiple **Best Paper / Editor’s Choice Awards**, including *Risk Analysis* (2021) and ASCE *Computing in Civil Engineering* (2022).

Mostafavi serves on editorial boards of four ASCE journals, reviews proposals for major U.S. and international funding agencies, and frequently briefs National Academies panels on infrastructure resilience.

Mostafavi’s work integrates community‑scale big data with advanced machine‑learning models to map and forecast infrastructure and social vulnerability. His group developed network‑level decision‑support tools now being adopted by the Texas Department of Transportation for statewide resilience planning, and methods accepted by the World Bank for road‑network risk assessments in Haiti, Colombia, and Guyana. Mostafavi is the author or coauthor of more than 450 journal and conference papers. His research focus is cutting-edge, interdisciplinary research at the nexus of urban science, complex networks, and artificial intelligence with special interest in use of novel data sources, such as human mobility or crowd-sourced social network reporting to collect real-time post-disaster urban conditions to prioritize rescue and recovery. Other research focuses on the applying machine language to large datasets to predicts infrastructure vulnerabilities.

Mostafavi’s UrbanResilience.AI Lab has released a family of open‑access analytics platforms—collectively branded Disaster AI—that provide near‑real‑time situational awareness for emergency managers and infrastructure owners:

• DamageCAT (2025): hierarchical transformer-U-Net model that classifies building damage into typologies from paired pre- and post-disaster satellite imagery, to deliver rapid, structured, and high-accuracy destruction mapping to sharpen post-disaster resource deployment^[3]
• Flood-DamageSense (2025): multimodal Mamba-based deep-learning system with multitask outputs (building damage state, floodwater extent, and building footprint) that fuses SAR/InSAR, optical basemaps, and a historical flood-risk layer to deliver actionable, building-scale flood damage maps within minutes—even under full cloud cover and minimal structural change.^[4]
• Evac-Cast (2025): interpretable XGBoost model (with SHAP explainability) that forecasts census-tract evacuation rates during hurricanes and wildfires using readily accessible hazard, vulnerability, readiness, and built-environment features, enabling survey-free, high-resolution predictions to support targeted emergency resource planning and evacuation decision-making.^[5]
• DisastIR (2025): a comprehensive information-retrieval benchmark tailored to disaster management, encompassing 9,600 realistic user queries and more than 1.3 million query–passage pairs across 48 retrieval tasks spanning six real-world search intents and eight major disaster categories with 301 event types, revealing wide model performance disparities and underscoring the need for domain-specific IR evaluation to guide effective decision-making during emergencies.^[6]
• Precipitation-Flood Depth Generative Pipeline (2025): generative machine-learning pipeline that uses a cell-wise depth estimator trained on limited physics-model outputs and CTGAN-generated synthetic precipitation–inundation events (10,000 samples) to produce high-resolution probabilistic flood maps, enabling scalable, time-efficient flood-risk modeling for urban planning and mitigation in data-scarce regions.^[7]
• FairMobi-Net (2025): fairness-aware deep learning model that predicts human mobility flows at fine spatial scales, designed to reduce systematic underprediction in underserved and vulnerable communities, thereby supporting more equitable disaster evacuation planning, urban mobility management, and resilience-oriented policy decisions.^[8]
• FloodDamageCast (2024): near‑real‑time flood‑damage nowcasting framework that blends GAN‑based data balancing with gradient‑boosted trees to map residential property‑damage severity at 500 m × 500 m resolution during unfolding events.^[9]
• Elev‑Vision (2024): computer‑vision pipeline that segments Google Street View imagery to infer each structure’s lowest‑floor elevation, a key input for depth‑damage and insurance models without costly field surveys.^[10]
• FloodGenome (2024): interpretable random‑forest model (with SHAP explainability) that decodes the hydrologic, topographic, and built‑environment “DNA” governing parcel‑level flood‑risk predisposition across U.S. metro areas.^[11]
• CrisisSense-LLM (2024): instruction fine-tuned open-source large language model adapted for multi-label social media text classification in crisis contexts, enabling simultaneous tagging of disaster event type, informativeness, and human-aid needs—significantly enhancing situational awareness from social sensing during emergencies.^[12]
• MaxFloodCast (2023): ensemble machine‑learning system trained on hydrodynamic simulations that delivers block‑level peak‑inundation‑depth forecasts within seconds, reducing physics‑based model runtimes from hours to seconds for emergency routing and flood‑plain management.^[13]
• Resili‑Net (2023): deep‑learning framework for community‑resilience rating that uses 12 socio‑technical features to classify census areas into five resilience tiers and reveals feature importance for targeted capacity‑building.^[14]
• FloodRisk‑Net (2023): unsupervised graph deep‑learning model that captures spatial dependencies and nonlinear hazard–exposure–vulnerability interactions to assign emergent flood‑risk levels across urban areas.^[15]
• DAHiTrA (2022): hierarchical‑transformer network that automatically delineates building footprints and classifies post‑disaster damage severity from high‑resolution satellite imagery within minutes of image availability.^[16]

Several of these tools have already been piloted by agencies such as the Texas Department of Transportation and the World Bank^[17] to inform infrastructure‑resilience planning and post‑disaster response.

• > 200 refereed journal articles; > 350 total publications.
• > 7,200 Google Scholar citations; h‑index = 45 (July 2025).
• Listed in the **Stanford–Elsevier Global Top 2% Scientists (Civil Engineering & AI; 2021‑2023).
• Ranked #181 of 226,271 civil‑engineering scholars worldwide (top 0.08%) by ScholarGPS.

• Disaster City Digital Twin^[18]
• Social Media for Intelligent Public Information and Warning in Disasters^[19]
• Social Sensing in Disaster City Digital Twin: Integrated Textual–Visual–Geo Framework for Situational Awareness during Built Environment Disruptions^[20]
• Resilience of infrastructure systems to sea-level rise in coastal areas: Impacts, adaptation measures, and implementation challenges^[21]
• Artificial intelligence for flood risk management: A comprehensive state-of-the-art review and future directions^[22]
• Social media for intelligent public information and warning in disasters: An interdisciplinary review ^[23]
• Equitable resilience in infrastructure systems: empirical assessment of disparities in hardship experiences in vulnerable populations during service disruptions,^[24][25]
• An integrated physical-social analysis of disrupted access to critical facilities and community service-loss tolerance in urban flooding^[26]
• Energy inequality in climate hazards: Empirical evidence of social and spatial disparities in managed and hazard-induced power outages^[27]
• Equitable resilience in infrastructure systems: empirical assessment of disparities in hardship experiences of vulnerable populations during service disruptions^[28]

In 2024 Mostafavi founded **Resilitx AI**, a spin‑off that commercializes his lab’s AI‑driven digital‑twin platform for disaster management and situational awareness. Resilitix products applies near-real-time data, including novel data sources; analytics; and AI-powered insights within a full-picture digital geographic model to provide dynamic status updates and predictive insights to first responders and emergency managers during and in the aftermath of disaster events. The startup has received an NSF SBIR Phase I award and a Texas A&M Innovation Award, and its technology was deployed during the 2024 Atlantic hurricane season.

After Hurricane Beryl made landfall in July 2024, Resilitix, using EmergenCITY Digital Twin technology, Resilitix delivered near-real-time situational awareness to local and state emergency responders. Resilitix monitored food lifeline downtime resulting from a power outage that affected more than 2 million people. Coupled with a heat wave, the power outage increased health risks. Resilitix monitored hospitals to identify areas with increased health risks to support proactive responses.

• The National Science Foundation publication Science Matters: Digital Twin observed that the bidirectional interaction between the physical and virtual realms drive the up-to-the-minute status updates.^[29]
• The National Science Foundation News: Resilitix supports Beryl emergency response efforts with NSF-supported digital twin technology^[30]
• Houston InnovationMap: Who’s Who: 3 Houston innovators to know this week^[31]
• Houston InnovationMap: (Podcast) Houston resilience tech innovator proves out platform amid Hurricane Beryl^[32]
• Houston InnovationMap: Stepping Up: Houston Startup Taps into Tech to Provide Key Data in the Wake of Hurricane Beryl^[33]

• Google Scholar
• Urban Resilience AI webpage
• Resilitix

• Disaster informatics
• Urban resilience
• Digital twin