GrammaTech will explore the application of machine learning and artificial intelligence to automate software tools' design, testing and implementation under a contract of an undisclosed sum from the Defense Advanced Research Projects Agency.
The company said Monday research findings will be available on Mnemosyne, its open source tool being designed to serve as an automated software development assistant that will help ensure documentation aligns with implementation.
With support from GrammaTech, DARPA aims to help developers employ statistical machine learning, formal methods and search-based software engineering to tackle problem definition and high-level software design via the Intent Defined Adaptive Software program.
"By separating problem definition from implementation, this project will help make it possible for software developers to focus on “˜what to do', by handling the details of “˜how to do it' using automated code synthesis and adaptation tools," said Alexey Loginov, vice president of research at GrammaTech.
The agency plans to roll out a framework that optimizes modern software development environments and yields various types of intent specification. DARPA also intends to develop synthesis tools such as test generation and automated repair platforms.
GrammaTech has tapped researchers from the University of Texas at Austin and the Massachusetts Institute of Technology to contribute to the study.