As a freelance AI Evaluator and Multimodal Content Reviewer, I’ve evaluated 3,500+ AI-generated outputs across images, video, audio, and conversational text, achieving 96% rubric alignment accuracy.
I score instruction following and reference preservation independently on R2I tasks, and use pairwise preference ranking to assess multimodal model outputs. My reviews address prompt fidelity, identity preservation, and aesthetic quality.
I’ve authored 800+ detailed critiques of LLM and diffusion model responses, identifying hallucinations, factual errors, and safety violations. I’ve also engineered and stress-tested 600+ evaluation prompts across creative domains and multi-turn dialogue.
In my Multi-Turn Multimodal Evaluation Portfolio project, I designed a 100-conversation evaluation set and documented scoring matrices and critiques in Notion. I also created a 200-prompt benchmark comparing DALL-E 3, Midjourney v6, and Stable Diffusion XL.
