Haorui Yu

PhD Candidate

About

I am a PhD candidate at the University of Dundee, DJCAD, working on vision-language models for cultural understanding in heritage and arts contexts.I build VULCA, an open-source visual AI evaluation and source-backed review workflow. The project maps visual assets into structured review fields, source context, evidence cards, and owner routes so AI-assisted creative work can be inspected before reuse.My current interests sit between visual AI, creative analytics, product engineering, and human-review workflows for AI-assisted visual assets. I am especially interested in scoped product / AI engineering internships or research engineering projects where careful software, structured data, and visual reasoning meet real product workflows.Technical focus: Python, TypeScript, multimodal AI evaluation, VLM workflows, MCP, ComfyUI, JSON/Markdown artifacts, GitHub Actions, PyPI packaging, and source-safe visual review processes.Project: https://github.com/vulca-org/vulcaGitHub: https://github.com/yha9806Google Scholar: https://scholar.google.com/citations?user=7vI0D4AAAAAJ&hl=en

Country

United Kingdom

City

Dundee

Industry

Information Technology & Services

Skill

Model Context Protocol, AI Evaluation, Artificial Intelligence (AI), Vision-Language Models, ComfyUI, GitHub Actions, TypeScript, Prompt Design, Python, 混合媒介艺术, 软件项目管理, AI, 数据挖掘, 艺术, 数字人文, 定量研究, 英语

Experience

PhD Candidate

University of Dundee

2024-9 - Present · 2 yrs 1 mo

邓迪

PhD candidate at the University of Dundee, DJCAD, researching vision-language models for cultural understanding in heritage and arts contexts. Building VULCA, an open-source visual AI evaluation and source-backed review workflow for cultural, design, media, and creative-analytics contexts. - Maps visual assets into source context, review fields, evidence cards, and owner routes. - Works across Python, TypeScript, MCP, ComfyUI, VLM evaluation pipelines, JSON/Markdown artifacts, and source-safe visual review processes. - Maintains open-source project structure, documentation, tests, GitHub workflows, and PyPI packaging. - Focused on product-facing workflows where structured data, visual reasoning, and human review meet real creative operations.

Education

University of Dundee

Vision-Language Models for Cultural Understanding in Heritage and Arts Contexts

2024-9 - 2027-6 · 2 yrs 10 mos

PhD candidate at Duncan of Jordanstone College of Art & Design (DJCAD). Research focus: multimodal AI evaluation, cultural understanding, visual reasoning, art and heritage contexts, and source-backed visual review workflows.