Software Engineering Institute (SEI) Applied Research
Work conducted under this program will enable resilient mission assurance in heterogeneous and contested environments through the verification and validation of system performance and architecture. The program will also assist the Department of War (DoW) in retaining a long-term advantage in the areas of software-intensive systems and cyber security by enhancing assurance, exploiting automation and Artificial Intelligence (AI), and understanding human-computer interaction. Software Engineering Institute (SEI) Applied Research has two main research thrusts with known military applications: (1) Software Engineering, Systems Verification and Validation, and Mission Assurance (formerly Mission Assurance); and (2) Information Assurance. This area is increasingly being applied to AI and autonomous systems.
FY2027 planned work Work toward hybridizing AI methods for software generation for different classes of software, cloud, and information systems so that real time assessment and updating of system architectures can be automated, verified and validated as part of the software Development and Operations lifecycle.
FY2026 to FY2027 change The Decrease of $1.689 million between FY 2026 and FY 2027 reflects a realignment to National Defense Education Program.
FY2026 plans — current year Leverage AI generated software methods to assess feasibility for systems composition. Enable extensions of verification and validation methods for code composition to system composition.
FY2025 accomplishments Innovate new methods for artificial intelligence (AI) generated software which enable integrated verification and validation methods for correctness of code. Hybridize different AI models including but not limited to graph neural networks, Bayesian convolutional networks, and transformers to enable validated software output. Leverage as many commercial and open-source approaches as possible.
FY2027 planned work Develop AI methods that enable graph-based verification and validation of software and test the methods on AI generated software and system design.
FY2026 to FY2027 change No change between FY 2026 to FY 2027 funding.
FY2026 plans — current year Develop methodologies to safeguard training data for automated code generation and vulnerability analysis and identify how such training sets can lower overall risk of errors and security vulnerabilities in software.
FY2025 accomplishments Assess how training methods in AI generated code and cyber vulnerability analysis can themselves by compromised with poor software examples or training data.