About Me

I am a PhD student at Queen Mary University of London and a DeepMind PhD Scholar, specializing in Vision-Language Foundation Models with an emphasis on fairness, safety, and multimodal reasoning. My work is driven by a commitment to ‘AI for Good,’ exploring innovative solutions in machine learning—particularly for medical applications—that balance cutting-edge performance with ethical and societal considerations. My research includes projects on transfer learning, computer vision, and graph neural networks, showcasing my dedication to leveraging technology for real-world impact.

I hold a Master’s Degree from the American University of Beirut, where I also served as a teaching assistant in programming and data structures. Additionally, I have collaborated with platforms like deeplearning.ai to develop and test new ML curricula, reflecting my passion for both advancing research and democratizing AI education.

Education

• PhD in Computer Science
  Queen Mary University of London
  Trustworthy Vision Language Models with focus on Fairness, Reasoning, and Safety
  DeepMind PhD Scholar
• Master’s in Electrical and Computer Engineering
  American University of Beirut
  Thesis:Transferability of Graph Neural Networks for Time Series Applications
  Research on transfer learning, computer vision, and graph neural networks

Research

• PhD Researcher
  Queen Mary University of London
  Working on Vision Language Foundation Models while focusing on fairness, safety and reasoning.
• Research Assistant
  American University of Beirut
  Assisted in various AI projects, particularly in transfer learning, computer vision, spatio-temporal graph neural networks, ai for agriculture, bias in hate speech detection systems, and deployment of ai models.
• Machine Learning Researcher
  Fields Institute and University of Toronto
  Proposed adding fractional derivatives and features (vaccination, mean society age ,stringency index) then applying PCA to Hurricane forecasting model.

Teaching

• Teaching Assistant
  University College London
  Supported courses in machine learning and programming, enhancing student understanding of complex concepts.
• Demonstrator
  Queen Mary University of London
  Assisted in teaching deep learning and computer vision courses, providing hands-on guidance and support to students.
• Teaching Assistant
  American University of Beirut
  Delivered courses in programming and data structures, emphasizing foundational education in computing principles.
• Instructor
  Udacity
  Led online instructional sessions focused on Python programming and introductory AI concepts, tailoring lessons to a diverse online audience as part of AWS AI scholarship and Google Launchpad scholarship.
• Teaching Assistant
  Data Science and AI in Health Summer School
  Directed sessions on AI applications in health, demonstrating the impact of machine learning in medical diagnostics through interactive learning.

Publications

• FairCoT: Enhancing fairness in text-to-image generation via chain of thought reasoning with multimodal large language models
  MLLM reasoning-based method for debiasing text-to-image generation.
  Read more
• Data matters most: Auditing social bias in contrastive vision–language models
  Audits social bias in contrastive vision–language models and analyzes the role of data in bias outcomes.
  Read more
• Breaking language barriers or reinforcing bias? A study of gender and racial disparities in multilingual CLIP
  Studies gender and racial disparities across multilingual CLIP-style models in 10 languages.
  Read more
• FairJudge: MLLM judging for social attributes and prompt image alignment
  Uses MLLMs to evaluate social attributes and prompt–image alignment for text-to-image outputs.
  Read more
• Towards deployment-centric multimodal AI beyond vision and language
  Discusses deployment-centric multimodal AI challenges and directions beyond vision–language settings.
  Read more
• Bias mitigation in hate speech detection
  Book chapter covering bias mitigation in hate speech detection.
  Read more
• Spatio-Temporal Graph Neural Networks: A Survey
  Survey and taxonomy of spatio-temporal GNN methods and applications.
  Read more
• The power of transfer learning in agricultural applications: AgriNet
  Presents AgriNet for transfer learning in agricultural applications (Frontiers in Plant Science).
  Read more

Awards

• DeepMind PhD Fellowship
  
• Winner of AI Hack Covid Hackathon
  
• LIRA Fund Award
  
• FYP Accelerator Award (Advisor)
  

Contact

• Email: z.alsahili@qmul.ac.uk
• LinkedIn: Zahraa Al Sahili
• GitHub: zahraaalsahili