M. Sonza Reorda
Papers
11
Total Citations
88
H-Index
6
About
M. Sonza Reorda is a leading researcher at the forefront of hardware reliability for artificial intelligence, specializing in the resilience of Graphics Processing Units (GPUs) and Neural Networks. His work critically addresses the challenge of ensuring that AI accelerators—particularly those powering Convolutional Neural Networks (CNNs) in safety-critical domains like automotive and robotics—remain dependable over lifespans exceeding ten years. Reorda has pioneered multi-level frameworks for injecting permanent faults into GPU architectures, including specialized Tensor Core Units (TCUs), to evaluate their impact on CNN accuracy and structural integrity. His research uniquely bridges reliability with security, exploring encryption mechanisms to protect intellectual property in non-volatile memories while enhancing fault resilience. With over 80 citations across his most influential papers, including a landmark 2022 study with 32 citations, Reorda has developed compaction methods for in-field GPU testing and assessed the reliability of split computing for IoT and 5G systems. His contributions are instrumental in shaping certification standards for AI in safety-critical applications, making him a pivotal figure in the quest for robust, long-lasting intelligent systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4Evaluating the Reliability of Supervised Compression for Split Computing8 citations · 2024
- 5A Compaction Method for STLs for GPU in-field test8 citations · 2022
- 6
- 7
- 8
- 9
- 10