
Detail-oriented and analytical AI Trainer with hands-on experience optimizing large language and multimodal models. Proven ability to evaluate complex audio and video data through side-by-side comparative analysis. Specialized in stress-testing AI models using advanced adversarial prompt engineering to identify flaws, improve system robustness, and enhance generation accuracy.
*Multimodal Evaluation: Conducted side-by-side comparative assessments of AI-generated audio and video outputs to evaluate quality, alignment, and human preference.
*Adversarial Testing: Designed and executed "red-teaming" strategies to purposely break AI models, identifying vulnerabilities in contextual understanding.
*Edge-Case Analysis: Prompted models using mismatched data, including sending conflicting images and incorrect visual pinpoints, to test system boundaries.
*Data Annotation: Labeled, rated, and categorized large datasets of visual and auditory content based on strict project taxonomy guidelines.
*Feedback Engineering: Authored detailed, objective rationale reports explaining model failures and ranking preferences to guide engineering teams in model optimization.