Postdoctoral Researcher

Department of Computer Science, University of Idaho

• Advisor: Xiaogang Ma.
• Super-resolution of very large scientific images: proposed InfScene-SR, which scales fixed-size diffusion super-resolution models to super-resolve arbitrary size images — variance-corrected joint denoising of overlapping patches eliminates the boundary artifacts of patch-based approaches at constant memory cost; demonstrated on remote-sensing imagery (preprint 2026).
• Reliable co-registration of large optical satellite imagery: proposed SCDF, a training-free estimator whose filters self-calibrate on each image pair; on a new 584-pair ground-truth benchmark, it registers every pair without failure and lowers the best baseline's median error from 6.83 to 4.17 m (preprint 2026; submitted to IEEE TGRS).
• Developing the official mobile application of mindat.org, the world's largest open mineralogy database, giving geologists offline field access to mineral occurrence data; now in internal testing.
• Early-stage work on agent systems that strengthen the information-retrieval capabilities of large language models for question answering, with evaluation targeting mineral-science data.
• Collaboration with NCEAS (UC Santa Barbara) on deep-learning species distribution models revealing hurricane-driven seabird displacement (Ecological Informatics 2026), and with NCEAS and The Nature Conservancy on continued invasive-species (iceplant) mapping begun during the Ph.D.

Graduate Research Assistant

Department of Computer Science, University of Idaho

• Interactive image segmentation: proposed CFR-ICL, a cascade-forward refinement model with iterative click loss that set the state of the art on standard benchmarks while requiring 33% fewer annotation clicks (AAAI 2024).
• Generative modeling: introduced a training-free diffusion fusion method that enables small text-to-image models to synthesize large-content images, lowering FID from 11.19 to 4.02 (AAAI 2025).
• Medical imaging: proposed MIRST-DM, an adversarial defense for ultrasound-based breast cancer classification that improves robustness to attacks by 45% (MICCAI 2022), and co-authored a benchmark study of breast ultrasound image classification (preprint 2023).
• Nuclear fuel characterization, in collaboration with Idaho National Laboratory (INL): developed a data-efficient instance-segmentation approach for fission gas bubbles in irradiated metallic fuel, reaching 93% recall from only 827 annotated instances (Scientific Reports 2023), and contributed to fission-gas pore segmentation and classification in U10Zr fuel and to TRISO fuel cross-section characterization (Materials Characterization 2024, 2026).
• Natural language processing for nuclear reliability, with INL: built an end-to-end pipeline for automated cause–effect extraction from licensee event reports — spanning PDF parsing, text annotation, and language-model fine-tuning — reaching 99.1% accuracy (preprint 2024), and contributed to failure-event mining of U.S. nuclear power plant reports with fine-tuned LLMs (Risk Analysis 2026).
• Research infrastructure: built and maintained two interactive-segmentation annotation platforms, Auto-IS (general-purpose) and SegMap (remote sensing; QGIS plugin and web), in use at INL, the University of Idaho, NCEAS (UC Santa Barbara), and The Nature Conservancy, underpinning 12+ interdisciplinary projects across materials science, medicine, biology, geography, and geology.
• Administered the group's GPU computing infrastructure (Linux servers with 15 RTX 8000/6000 GPUs) supporting 20+ researchers.

Research Assistant

Guangzhou Institute of Geography

• Implemented a mechanistic, remote-sensing-driven leaf turnover model reproducing foliar area and photosynthetic seasonality at four Amazon evergreen forest sites, and integrated it into ORCHIDEE — a core component of the IPSL Earth System Model contributing to CMIP and IPCC assessments — in collaboration with the model's developers at the Institut Pierre-Simon Laplace (IPSL, France).
• Analyzed time-series and spatial datasets to characterize relationships between plant phenology and ecological indices.
• Established a distributed high-performance computing cluster enabling global-scale model simulations, and developed parallel Python toolkits that scaled remote-sensing image analysis from the gigabyte to the terabyte range.

Python Web Developer

Yaran / Mostfun / Maizi

• Built Python backends for three companies: microservices and CI/CD for mobile-game servers, a 3D-model sharing platform with single sign-on, and full-stack web apps (Django, Tornado, Sanic, MySQL, Cassandra, Docker, Terraform).

Ph.D. in Computer Science

University of Idaho

Graduate Studies in Technology Management

University of Idaho

B.S. in Applied Physics

Chengdu University of Technology

Outstanding Ph.D. Student

University of Idaho

MIDA Star Award

Machine Intelligence and Data Analytics Lab, University of Idaho

Outstanding Graduate Student in Idaho Falls

University of Idaho

Meritorious Winner, Interdisciplinary Contest in Modeling (ICM)

COMAP

Proposal Review Panelist, National Science Foundation (NSF)

Reviewer, AAAI Conference on Artificial Intelligence (AAAI)

Reviewer, Applied Computing and Geosciences

Reviewer, European Conference on Computer Vision (ECCV)

Reviewer, Formal Ontology in Information Systems (FOIS)

Reviewer, International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI)

Reviewer, IEEE International Symposium on Biomedical Imaging (ISBI)

Reviewer, The Web Conference (WWW)

InfScene-SR: Spatially Continuous Inference for Arbitrary-Scene Image Super-Resolution

Invited lecture, CS 5621 Data Science, University of Idaho · Moscow, ID

Guided and Variance-Corrected Fusion with One-Shot Style Alignment for Large-Content Image Generation

AAAI Conference on Artificial Intelligence (AAAI-25) · Philadelphia, PA

Poster

CFR-ICL: Cascade-Forward Refinement with Iterative Click Loss for Interactive Image Segmentation

CS 501 Graduate Seminar, University of Idaho · Idaho Falls, ID

CFR-ICL: Cascade-Forward Refinement with Iterative Click Loss for Interactive Image Segmentation

AAAI Conference on Artificial Intelligence (AAAI-24) · Vancouver, BC, Canada

Poster

MIRST-DM: Multi-Instance RST with Drop-Max Layer for Robust Classification of Breast Cancer

CS 501 Graduate Seminar, University of Idaho · Idaho Falls, ID

MIRST-DM: Multi-Instance RST with Drop-Max Layer for Robust Classification of Breast Cancer

Guest lecture, CS 504 Adversarial Machine Learning, University of Idaho · Idaho Falls, ID

Adversarial Attack and Defense

Guest lecture, CS 504 Adversarial Machine Learning, University of Idaho · Idaho Falls, ID

MIRST-DM: Multi-Instance RST with Drop-Max Layer for Robust Classification of Breast Cancer

International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2022) · Virtual

Poster

CS 577 — Python for Machine Learning

University of Idaho

Mentored graduate students on applied machine-learning course projects and provided detailed assignment feedback.

CS 574 — Deep Learning

University of Idaho

Guided graduate students through deep-learning course projects and provided detailed feedback.