Takumi Uezono

Papers

Total Citations

H-Index

About

Takumi Uezono is a leading researcher at the intersection of hardware reliability and artificial intelligence, specializing in functional safety for AI-enabled systems. His work addresses a critical challenge: ensuring that deep learning accelerators and FPGA-based control systems remain trustworthy even when deployed in safety-critical environments like robotics, healthcare, and autonomous mobility. Uezono’s most influential contribution is his 2021 paper on efficient functional in-field self-test for deep learning accelerators (17 citations), which pioneered techniques to generate high-quality tests that run during normal operation—a vital capability for meeting safety and reliability requirements without system downtime. He further advanced the field with his 2024 work on real-time diagnostic techniques for AI-enabled systems (4 citations), addressing the growing need for continuous health monitoring in neural network deployments. Earlier, Uezono demonstrated his expertise in FPGA reliability through his 2017 paper on soft error correction via interruption scrubbing (2 citations), introducing a high-speed CRAM diagnosis macro that detects and corrects neutron-induced errors in SRAM-based FPGAs. His research is essential reading for engineers and researchers developing dependable AI hardware for real-world applications.