Federal Project Grant Award Summary

The National Science Foundation's Division of Information and Intelligent Systems awarded $650,000 to the University of California Irvine under the Computer and Information Science and Engineering program (CFDA 47.070) for the project "Design Automation of Resilient and Interpretable Neural Architectures for Safety-Critical Systems" (Award Date: July 1, 2026; Completion Date: June 30, 2029). The project delivers research methods and design tools for developing neural network models that are both resilient to faults and interpretable for safety-critical applications, including autonomous driving and automated insulin delivery systems. Key deliverables include novel architecture search frameworks that automatically generate optimized neural network designs balancing accuracy, robustness, and computational efficiency; formal verification techniques to guarantee safety properties; and runtime monitoring systems to detect abnormal behavior during operation. The research team also develops a simulation-based testing platform for evaluating system performance under realistic conditions including sensor noise, hardware faults, and rare scenarios.

The project produces a comprehensive toolkit and design principles intended to enable the development of artificial intelligence systems that operate safely and predictably in high-stakes environments. By advancing techniques for detecting and responding to unexpected conditions, this work aims to reduce failures in autonomous systems, enhance public trust in emerging technologies, and contribute to national priorities in transportation safety, healthcare advancement, and trustworthy artificial intelligence workforce development. The research addresses critical gaps in AI reliability and transparency for modern safety-critical systems that increasingly depend on artificial intelligence for consequential decision-making.