Summary
Youmna Farag is a research scientist at Toshiba Research Europe with a decade of experience bridging academic NLP research and practical software engineering. She focuses on embodied AI and language-and-vision agents, currently building task-aware agents in the AI2-THOR simulator using LLMs and computer vision. Her PhD from Cambridge developed neural methods for discourse coherence and an evaluation framework probing what coherence models learn, with applications to everyday writing and education. Previously she led mobile and web engineering teams in Egypt, giving her a strong product-oriented perspective on deploying ML systems. Her toolkit spans transformers, recurrent and convolutional networks, adversarial and multi-task training, and deep-learning interpretability across Theano, Keras, and PyTorch. Notably, she combines deep theory on discourse with hands-on agent development, making her work both scientifically rigorous and directly applicable.
10 years of coding experience
10 years of employment as a software developer
PhD, Computer Sciencce, PhD, Computer Sciencce at University of Cambridge
Bachelor of Science (B.Sc.), Computer Science, 3.89, Bachelor of Science (B.Sc.), Computer Science, 3.89 at The American University in Cairo
Arabic, English