Real-world AI Lab (RAIL)
Physics-informed AI & Hydrology
We are the Real-world AI Lab (RAIL), a cross-disciplinary research group bringing together experts from diverse professional domains to develop AI technologies with real-world impact. Our mission is to bridge cutting-edge AI research with practical, deployable solutions across industries and society.
Latest News
May 7, 2026
Google TPU Grant Awarded.
for DualFloodGNN-TPU: A Scalable Physics-Informed Graph Foundation Model for Large-Scale Hydrodynamic Flood Simulation.
Go to News Page
Mar 16, 2026
UrbanFloodBench: Flood Modelling - Winners Announcement
Completed on 16 March 2026. We're excited to announce the Top 5 winners of UrbanFloodBench: Flood Modelling 2026. Champion: Jobayer Hossain 1st Runner Up: SGD Lai Lai 2nd Runner Up: Yukiya Tsukada 4th Place: Shrey Gandhi 5th Place: mtmr_s1 We are incredibly inspired by the level of participation and creativity demonstrated throughout the competition. A total of 292 participants submitted 3,799 entries, reflecting strong global engagement. Based on voluntary responses, participants represented over 20 countries, highlighting the competition's international reach. All top solutions passed validation for reproducibility, originality, and performance. From novel modeling approaches to highly optimized solutions, participants pushed the boundaries of AI-driven flood prediction and hydraulic modeling. Huge congratulations to everyone who participated! Thank you for being part of this competition. We hope you found the experience rewarding and that it helped you deepen your skills, explore new ideas, and connect with the community. Your enthusiasm made this event a success, and we look forward to seeing you in future competitions!
Go to News Page
Dec 18, 2025
UrbanFloodBench: Flood Modelling
Launched in January 2026. This competition aims to advance urban flood forecasting in coupled 1D-2D hydraulic systems, where underground drainage networks interact dynamically with surface overland flow. Participants are challenged to predict water levels at both 1D drainage nodes and 2D surface nodes, using rainfall and hydraulic states.
Go to News Page