Diana Abagyan is a Member of Technical Staff at Cohere, specializing in pretraining data for large language models. Her expertise lies in machine learning and natural language processing, with a focus on efficient, multilingual systems and low-resource languages. She led the pretraining and tokenization efforts for the Tiny Aya project and has released work on efficient multilingual tokenization techniques. Diana strives to improve the accessibility and performance of language technologies across diverse linguistic settings.