AI Video Model Evaluator & Data Trainer | Verity Labs (Anuttacon) (Video Evals team)
VideoEvaluated AI-generated video outputs using structured, multi-category rubrics across visual, identity, audio, and composite quality dimensions. Applied calibrated severity and duration taxonomies and reproducible scoring scales to align ratings across a distributed annotation team. Conducted comparative Unilateral GSB (Good/Same/Bad) head-to-head ranking to support preference-based model preference training. • Rated Dynamic Effect metrics including structural accuracy, motion stability, and natural vividness. • Checked ID Consistency (Face ID, Body ID, and screen-content fidelity to reference images) and Audio Matching (lip-sync accuracy and emotion alignment). • Assessed Color Consistency and combined Face/Hands/Body/Movement (FHBM) quality using taxonomy-driven criteria. • Participated in calibration syncs to maintain scoring alignment and update guidelines as the taxonomy evolved.