AI Trainer – Anuttacon (June 2025 – Present)
Labeled and evaluated multimodal in-game content and AI-generated video outputs using RLHF/SFT methodologies and project-specific rubrics. Assessed visual quality, coherence, and output refinement while performing comparative assessment and issue identification across generated content. Evaluated audio-visual alignment including audio synchronization, lip-sync accuracy, and emotion alignment. • Annotated images and videos as part of dataset building for model improvement • Applied RLHF/SFT evaluation criteria to rate and refine AI-generated video responses • Performed multimodal evaluation for synchronization, lip-sync, and emotional congruence • Supported LLM evaluation by scoring factuality, tone, contextual relevance, safety, and response quality