Muhammed Saeed

Muhammed Saeed develops AI systems that work for everyone, not just English speakers. Currently pursuing cutting-edge research in dataset curation algorithms, he focuses on training smaller, more efficient foundation models that rival larger ones while maintaining multilingual capabilities.
At Saarland University, where he earned his M.Sc. in Data Science and AI under Prof. Demberg with Eliza scholarship support, Muhammed pioneered discourse parsing for Nigerian Pidgin—a language spoken by 100 million people but previously ignored by NLP tools. His COLING 2025 paper achieved breakthrough results with 33.98% F₁ improvements, demonstrating how to extend advanced NLP to underrepresented languages.
His research portfolio spans top venues (ACL, COLING, Interspeech) and addresses critical challenges: multilingual machine translation, cultural bias evaluation in LLMs and text-to-image models, and semantic deduplication for multimodal data. During his MBZUAI internship, he advanced responsible AI practices, consistently working toward democratizing access to high-quality AI across linguistic and cultural boundaries.