James Ghawaly, “Low-power AI for Source Detection at the Edge”

This talk presents recent progress in the development of low-power AI algorithms tailored for source detection in low signal-to-noise ratio (SNR) environments. We will explore compact deep learning models, such as the Autoencoder Radiation Anomaly Detection (ARAD) approach, which enable unsupervised anomaly detection on resource-constrained hardware like single-board computers. The presentation will then highlight emerging techniques in neuromorphic computing that offer ultra-low power consumption, making them ideal for persistent, long-term radiation monitoring. We will conclude with a discussion on ongoing efforts and future research directions in low-power, AI-driven sensing systems.

Dr. James Ghawaly is an Assistant Professor of Computer Science & Engineering at Louisiana State University (LSU), where he researches modern and emerging machine learning techniques, with applications to national security, especially cybersecurity. He also holds joint appointments with the LSU Center for Computation and Technology (CCT) and the LSU Office of Academic Affairs (OAA). Prior to LSU, James was a staff research data scientist at Oak Ridge National Laboratory (ORNL), where he led research efforts to develop applied deep learning and neuromorphic computing approaches for federal sponsors. Recently, his work has been centered around transformer technology and Large Language Model (LLM) research. Throughout his education and career, James has developed a strong passion for neuromorphic computing, applied deep learning, and machine learning-enabled sensing.