Research

Human-centered urban AI for evidence, experience, and action.

I develop computational approaches that turn diverse urban data into accountable evidence for planning. My work follows a connected path: first understand what urban data can reveal, then explain how environments shape experience, and finally support decisions that affect people and places.

A connected research program

From sensing to explanation to action

01 · Selected papers

Sense cities and audit the data

I turn street-level imagery and other spatial records into interpretable urban evidence, while asking whose environments those data actually represent.

Workflow for measuring visual environmental quality from time-series street imagery.
Cities · 2024First author

Measuring visual quality over time

A deep-learning workflow measures six dimensions of visual environmental quality across public-housing neighborhoods in Singapore and New York over a decade.

Why it matters. Shows how repeated street imagery can support longitudinal, spatially comparable evidence for neighborhood equity and monitoring.
Time-series Street ViewDeep learningSpatial equity
Read article
Global Streetscapes dataset illustration.
ISPRS Journal of Photogrammetry and Remote Sensing · 2024Co-author

Building a global streetscape resource

Global Streetscapes provides 10 million crowdsourced street-level images from 688 cities, enriched with geographic, temporal, semantic, and perceptual attributes.

Why it matters. Creates open spatial data infrastructure for comparative urban research and for evaluating computer-vision models under varied real-world conditions.
Open spatial dataData enrichmentComputer vision
Read article
GSVantage pipeline linking Street View coverage with socioeconomic indicators.
ACM SIGSPATIAL GeoIndustry Workshop · 2025First author

Auditing who is represented in Street View

An open workflow evaluates the recency and coverage of Google Street View across four U.S. cities and relates image staleness to neighborhood characteristics.

Why it matters. Makes coverage bias visible before street imagery is treated as evidence about neighborhoods or populations.
Google Street View metadataACS + OpenStreetMapCoverage audit
Read paper