By cutting data layers, researchers create fast, explainable landslide detection AI

Arsalaan Ahmad, a 23-year-old computer science graduate from Cardiff University, has been named as the lead author of a research paper in Frontiers in Remote Sensing. The paper, co-authored by his mentors Dr. Oktay Karakuş and Professor Paul Rosin, explores the use of satellite AI to map natural disasters such as landslides. The team developed a framework that reduces the data layers in AI models from up to 30 to just eight, creating a compact model that is fast, inexpensive, and easy to audit.

This innovation allows disaster AI to run on cheaper hardware and delivers fast, explainable maps to emergency teams. Ahmad's journey to becoming a first-author researcher began with an on-campus internship proposal, which was accepted after he met with Professor Rosin. Despite initial skepticism from some professors, the project was ultimately successful, and Ahmad's hard work paid off with a publication in a top-tier academic journal.

Ahmad's experience highlights the importance of being proactive in seeking out research opportunities and not being discouraged by setbacks. He now plans to use his skills in the startup market, focusing on energy and climate technology.