CalypsoAI is helping the U.S. Navy accelerate AI adoption by setting with service’s first AI validation, verification, and robustness (VV&R) standards.
CalypsoAI recently completed a Small Business Innovation Research Program Phase 1 project with the US Navy. During the course of our Phase 1 effort, our team adapted our commercially available Vespr Testing Framework (VTF) to provide certification and testing standards for US Navy machine learning and artificial intelligence (AI/ML) applications.
The goal of this effort was to provide the Navy with a coherent framework that can be used to comprehensively understand AI/ML failure. This includes internal failure (failure due to poor development/debugging practices) and external failures (failure due to adversarial impact). For the US Navy, we call this new AI certification framework the Navy Validation, Verification, and Robustness (VV&R) methodology, setting it apart from traditional methods of software validation and verification (V&V). VV&R sets defined, quantitative standards around AI quality, robustness, and security while defining a repeatable, software-enabled process for ongoing VV&R efforts.