Ph.D. Student, Department of Geography and the Environment, The University of Texas at Austin
Looking for 2027 summer internshipBio: I am a first-year Ph.D. student in Geospatial AI at UT Austin, advised by Prof. Gengchen Mai in the SEAI Lab. I received my B.Eng. in Spatial Informatics and Digital Technology from the School of Remote Sensing and Information Engineering at Wuhan University in 2026, in the university's selective Experimental Class of Spatial Information.
I study geographic bias in GeoAI models and geospatial foundation models: how to measure it, understand where it comes from, and reduce it. My current work builds a statistics-based framework for quantifying and debiasing geographic bias, and extends it to the pre-training stage of geospatial foundation models. Before that, I worked on remote sensing change detection, which led to a first-author paper in IEEE J-STARS and a Chinese patent.
I was also a Google Summer of Code 2025 contributor at OSGeo, where I worked on graph reordering for pgRouting.
Email: wifibk [at] utexas [dot] edu
Making geospatial AI fair, measurable, and trustworthy.
AI models often perform systematically differently across places. Existing bias measures are tied to specific models and hide which spatial factors cause the gap. We treat model performance on Earth's surface as a spatial point pattern and use information theory to build model-agnostic, spatially explicit bias scores covering multi-scale heterogeneity, distance decay, and anisotropy.
Bias is easier to fix before it is baked in. I am developing a debiasing framework for GeoAI and foundation models, extending it to the pre-training stage of geospatial foundation models.
As an undergraduate, I developed content-aware enhancement methods for change detection, leading to CFNet (IEEE J-STARS) and a Chinese patent, and collaborated on ISANet for cropland change detection (IEEE ICCA).
Patent
Integrated the King and minimum-degree ordering algorithms from the Boost Graph Library into the pgRouting library.
Proposed a content-aware feature-enhancement network for remote sensing change detection, published in IEEE J-STARS (2025).
Conference reviewer
Activities