At Meta, I tested large language models and AI-driven communication systems for bias, misinformation, adversarial prompts, and manipulative language patterns. I used linguistic analysis and red teaming to uncover vulnerabilities across languages and cultural contexts.
I also researched attacks such as prompt injection, model inversion, and jailbreaks, collaborating with AI and security engineers, data scientists, developers, and ethicists on model robustness and security practices.
At TikTok, I led an emergency triage team reviewing livestreams and other content, and guided sensitive investigations and emergency escalations. We worked across teams to improve policies and processes while evaluating risk, threats, privacy, and user safety.
Earlier, I coached and mentored content moderators at Meta, supervised grant programs at Goodwill, and worked in media analysis and cultural advising with the US Army. My work has included Arabic-language media analysis, threat analysis, and English- and Arabic-language reporting.
