Haorui Yu

Meet Haorui

1 following

I help teams get AI output to a releasable state, with evidence: independent checks, one bounded repair, a human release decision, and a delivery record. PhD researcher in vision-language model evaluation (Dundee) and architect of VULCA, an open-source evaluation and control SDK (CLI, Python, MCP, ComfyUI). EMNLP 2025 Findings, WiNLP 2025, two 2026 preprints. Python, PyTorch, TypeScript. UK-based.