AI Training Specialist — Handshake AI Fellowship (Remote)
Evaluated and annotated AI-generated images using quality dimensions such as composition, realism, artifact detection, and prompt adherence. The work produced preference data used to support RLHF pipelines for model fine-tuning. This role focused on consistent assessment criteria and accurate labeling to improve downstream model performance.• Assessed image quality across multiple rubric dimensions including realism and artifacts.• Verified prompt adherence by checking alignment between inputs and generated outputs.• Contributed preference/ratings data to RLHF data preparation workflows.• Supported model improvement objectives through careful documentation of label outcomes.