Perkins graduated as valedictorian with a BS in Applied & Computational Mathematics (ACME) from Brigham Young University.

The growing frequency of mass shootings in the United States highlights the urgent need for systems that can guide individuals to safety in real time. Perkins’ research develops a framework that leverages graph neural networks and multi-agent deep reinforcement learning to dynamically generate optimal evacuation routes that minimize exposure to active threats. To enhance the model’s adaptability across diverse environments, he’s leveraging HPC resources to train it on thousands of randomly generated simulations. A complementary component of this work focuses on computer vision algorithms for real-time detection and tracking of active shooters. Perkins is developing Re-ID models capable of maintaining consistent tracking across multiple cameras with varying viewpoints. Additionally, he will enhance model robustness by developing image inpainting techniques to synthetically reconstruct occluded regions.