I conduct research at the intersection of artificial intelligence (AI) and healthcare. Specifically, I focus on developing AI systems that are (1) resource-efficient, requiring less data and supervision by medical professionals, and (2) trustworthy, exhibiting robust and fair performance to allow for their responsible deployment in a clinical setting.
I am currently on the research faculty in the Department of Urology at Cedars-Sinai. Most recently, I was a postdoctoral scholar at Caltech working on AI systems for surgical applications. I completed my PhD at the Computational Health Informatics Lab., at the University of Oxford. There, I designed deep learning algorithms that do 'more with less' in the context of cardiac data. Before this, I completed my undergraduate degree in biomedical engineering at The Johns Hopkins University. In the past couple of years, I have also had the privilege to work alongside colleagues and mentors at Ford Motor Company, the Mayo Clinic, Merck & Co., Flatiron Health, and Vicarious Surgical.
I also strongly believe in cultural preservation. To that end, I have (a) open-sourced an image dataset of Arab heritage for use by machine learning researchers, (b) created an online demo which allows users to identify UNESCO heritage sites in the Arab world based on an uploaded image, and (c) founded a non-profit organization, Arabs in Machine Learning (أمل), with a focus on engaging and amplyifying the voices of Arab researchers.
As a postdoctoral scholar, I designed deep learning algorithms that exploit intraoperative videos to quantify the core elements of surgery (think: what and how surgery is performed). In doing so, we can provide surgeons with feedback about their performance as a means to modulate their behaviour.
In the summer of 2021, I interned at Flatiron Health where I designed a natural language processing system that infers a clinical variable based on oncology patient visits. In the process, I also developed a framework for evaluating machine learning models in the absence of ground-truth labels.
Throughout my PhD, I focused on designing clinical deep learning algorithms that are less dependent on data, labels, and medical supervision. This involved leveraging generative modelling, self-supervised learning, and continual learning.
In the summer of 2020, I was fortunate to work alongside Antong Chen at Merck & Co. where I designed a meta-learning framework that leverages cardiac MRI data across medical centres in order to delineate the structures of the heart (segmentation) in a data-efficient manner. Our work can be found here.
In the summer of 2019, I had the pleasure of working with Kenneth Fetterly and Zachi Attia within the Department of Cardiovascular Medicine at the Mayo Clinic where I implemented self-supervised algorithms for coronary angiograms (read: X-ray videos of the arteries surrounding the heart).
In the summer of 2017, I interned at the Machine Learning Center of Excellence at Ford Motor Company, under the supervision of Kurt Godden and K.P. Unnikrishnan, where I used diagnostic trouble codes from vehicles to predict whether such vehicles will malfunction in the near future.
In the summer of 2016, I performed research at the Musculoskeletal Lab at Imperial College London, under the supervision of Alison McGregor, to identify how pressure is distributed within lower-limb prosthetic cuffs worn by amputee patients.
During my undergraduate studies, I conducted research at the Neuroengineering Lab. at The Johns Hopkins University to design and manufacture an upper-limb prosthetic device for data collection purposes.
I also led a team of engineers in the design of a surgical device that repositions patients during time-sensitive intra-operative complications. You can find an article on that here.
Feel free to reach out if you would like to chat.