H O

Senior AI Evaluator specializing in LLM evaluation, Japanese ASR validation, multimodal assessment, and translation QA. My work focuses on identifying hallucinations, reasoning failures, instruction-following issues, and semantic inconsistencies to improve AI quality, reliability, and auditability.

I have completed 1,500+ multimodal AI evaluations, 1,200+ Japanese ASR validation tasks, and 730,000+ words of technical translation and localization. Using structured rubrics (30+ evaluation criteria), I assess semantic fidelity, prompt alignment, safety, visual consistency, and overall model quality while producing structured English rationales for reviewer alignment and traceable QA.

I specialize in reproducible AI evaluation through structured decision frameworks, Red Flag detection, pattern-based QA, and YAML/JSON logging to support scalable Human-in-the-Loop (HITL) workflows. My experience spans multimodal evaluation, LLM assessment, translation QA, MTPE, and cross-lingual quality assurance across enterprise AI projects.

Beyond day-to-day evaluation, I develop reusable quality frameworks including Kernel Core, Translation OS Runtime, ModelRefiner, and Unified Cognitive OS to improve evaluation consistency, traceability, and governance across AI systems.