AI Design Curator
Contributing to the development and optimization of AI-powered design models by evaluating and labeling AI-generated visual outputs. The project focuses on improving model performance in UI/UX design understanding, ensuring outputs align with human-centered design principles, usability standards, and real-world design expectations. Data Labeling Tasks Performed Conduct rubric-based evaluation of AI-generated UI layouts using predefined criteria such as: Usability & accessibility Visual hierarchy & spacing Alignment and consistency Aesthetic quality and design coherence Annotate outputs using Label Studio, providing: Structured ratings (quantitative scoring) Detailed qualitative feedback for model improvement Identify design flaws such as: Poor navigation flow Inconsistent typography or spacing Misaligned components or weak visual hierarchy Compare multiple AI-generated variations and select the best-performing design based on human UX judgment Project Size & Scale Evaluated and annotated hundreds to thousands of design samples across multiple UI categories (e.g., dashboards, onboarding flows, mobile interfaces) Worked in iterative training cycles, contributing to continuous model refinement Collaborated in a cross-functional environment with AI engineers, product managers, and design researchers Quality Measures & Standards Maintained high annotation consistency by strictly adhering to defined evaluation rubrics and guidelines Followed inter-annotator agreement practices to ensure reliability and reduce subjectivity Performed self-audits and peer reviews to validate annotation accuracy Ensured low error rates and high precision in labeling to directly impact model performance Provided actionable feedback loops to improve both annotation guidelines and AI output quality Impact Helped improve AI model’s ability to generate user-friendly and aesthetically accurate UI designs Enabled systems to better replicate human design judgment and decision-making Contributed to enhancing end-user acceptance and usability of AI-generated interfaces