Urban representations
Multimodal and spatial models for imagery, maps, mobility, text, and city-scale signals.
Experimental urban intelligence
Build. Test. Question.
An open experimentation layer for urban AI models, agents, benchmarks, and prototypes grounded in real city questions.
Enter the lab$ initialize urban_system
input: spatial + visual + social data
task: understand city dynamics
constraint: validity, equity, transparency
output: testable evidence
Experiment stack
Urban AI as a testable system.
Projects will expose assumptions, training data, evaluation tasks, failure modes, and intended uses—not only polished demos.
Research agents
Experimental agents for urban analysis, evidence synthesis, scenario exploration, and planning workflows.
Grounded evaluation
Open tasks and leaderboards that test generalization, fairness, uncertainty, and practical usefulness.
Operating principles
Prototype responsibly.
Urban question firstTechnology follows a defensible problem and intended user.
Evaluation before spectacleDemos are paired with transparent tests and limitations.
Cross-city validityClaims distinguish local performance from generalizable evidence.
Open artifactsRelease code, task definitions, model cards, and reproducible results where possible.
Coming online
Follow the experiments.
Models, agents, benchmarks, and demos will be released through the public repository.