Multimodal AI Training Contributor & Language Expert | Freelance
Contributed to training LLMs on complex text, voice, and business datasets by performing RLHF and validation workflows. Conducted side-by-side model comparisons to reduce class bias and cultural or historical inaccuracies in model outputs. Performed red-teaming and jailbreak testing on Arabic models to improve safety and contextual reliability. • RLHF and output validation • Bias and cultural misalignment reduction • Safety/red-teaming and jailbreak testing • Model evaluation for marketing/business strategy alignment