Hello, I’m Xuhui.

I am an Applied Scientist and Machine Learning Engineer who enjoys working where large-scale AI systems, multimodal learning, and ML infrastructure meet. At Treverse LLC, I build production systems from end to end. My work includes a Go-based multilingual translation service deployed on AWS, an internal AI/ML platform for standardized training and CI/CD pipelines, an edge-first vision presort pipeline for high-speed barcode decoding, and recommendation models that produce personalized item suggestions for hundreds of thousands of users.

Before Treverse LLC, I explored several AI research domains at Vanderbilt University. At the AI Negotiation Lab, I led development of an LLM pipeline that automatically codes negotiation transcripts. It reduced annotation costs by more than 99% while increasing human–AI agreement from 30% to 80%. I presented this work at the 2025 AI Negotiation Summit at Harvard and MIT. At the Network and Data Science Lab, I studied multimodal learning across graph, vision, and language representations, including a redesign of LLaVA-style architectures that improved training efficiency while maintaining competitive performance.

I also enjoyed taking computer vision into an unexpected domain: archaeological analysis. As a Professional Research Assistant, I used Vision Transformers to improve ancient mortar classification accuracy from 60% to 97%, and developed a vision-only provenance algorithm that reduced similarity-analysis time from months to minutes. Earlier, as an Algorithm Engineer, I deployed computer vision storage-detection systems across approximately 30 warehouses and improved robotic localization pipelines for automated guided vehicles.

I hold an MSc in Data Science from Vanderbilt University with a 3.98/4.0 GPA and a BSc in Data Science with First Class Honours and Highest Distinction from Beijing Normal-Hong Kong Baptist University.

"Any sufficiently advanced technology is indistinguishable from magic."

Arthur C. Clarke, Clarke's Third Law