This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility: Publishing original research in a top-tier academic journal has been an ambition of Arsalaan Ahmad's since he enrolled in an undergraduate computer science program at Cardiff University. Among the self-confessed space nerd's achievements are teaching himself machine learning and amassing 50 million visits for a Roblox game he developed.
Now, having recently graduated from the School of Computer Science and Informatics, Ahmad is reflecting on becoming the lead author of a research paper in Frontiers in Remote Sensing. "The first thing I did was print the paper out so I could keep it in my room alongside the poster I had made about our findings," said Ahmad, who earned a BSc in computer science. The 23-year-old was born and raised in Oman, where the geography inspired the idea behind the research paper.
"I developed a real interest in landslides living among the mountains," Ahmad recalled. "In Oman, authorities take extensive measures during road and settlement construction to prevent natural disasters, which are increasing because of the effects of climate change." The paper, co-authored by Ahmad's mentors Dr. Oktay Karakuş and Professor Paul Rosin, examines the use of satellite AI to map natural disasters such as landslides.
The team explored a "black box problem" in which current AI models are bloated by up to 30 layers of data, making them slow, expensive and impossible to audit. Instead, they developed a framework that strips away redundant data to create a compact, 8-channel AI model that matches the accuracy of massive systems but runs significantly faster. This allows life-saving disaster AI to run on cheaper hardware, the researchers say.
Their solution also runs in resource-constrained environments, delivering fast, explainable maps to emergency teams when every second counts. The opportunity to make his idea a reality came when Ahmad applied for an on-campus internship. "I reached out to a few professors about the idea to work on something related to landslide segmentation, but most of them were pretty skeptical," he said.
"Professor Paul Rosin, who was our module leader for Computational Math, thought the idea was interesting, and he encouraged me to write an internship proposal. "I took a lot of time over it, researching different angles we could approach, and met with Paul to see what might work best. Eventually, the proposal was accepted, and I got to take full ownership of the project from beginning to end." Rosin, a professor of computer vision at Cardiff University's School of Computer Science and Informatics and one of the paper's co-authors, said, "The ideal outcome of our on-campus internships is to create something which might be publishable but, in my experience, it is rarely achieved.
Extract — continue reading at the source.
