Scalable Platform for Enterprise Engineering and Deployment towards Mathematics for the Discovery of Algorithms and Architectures (SPEED DIAL)

The Scalable Platform for Enterprise Engineering and Deployment towards Mathematics for the Discovery of Algorithms and Architectures (SPEED DIAL) is a Small Business Technology Transfer (STTR) solicitation issued by the Defense Advanced Research Projects Agency (DARPA) under the Department of Defense. The program aims to transition AI-driven algorithm discovery from research environments into practical commodities by creating a scalable platform that allows domain experts, such as physicists and engineers, to collaboratively discover, refine, and deploy high-performance algorithms for real-world applications. The project focuses on developing a closed-loop pipeline utilizing Monte Carlo Tree Search and Genetic Programming to autonomously generate and evaluate mathematical variations in the background, ensuring that optimal algorithms can be generated on-the-fly to meet specific engineering constraints. The 24-month period of performance is structured around several key milestones, beginning with the delivery of a system architecture and initial algorithm library by month 4, followed by the prototype discovery engine at month 8, and the release of API wrappers and auto-compilation tools by month 12. The project culminates in two Department of War use case demonstrations, including a target of over 10 percent improvement in computational efficiency for problems such as hypersonic design, and a final software release with a Phase III transition plan. Award selections are based on overall technical value, technical merit, investigator qualifications, and commercial potential. Compliance requirements include CMMC Level 2 certification for firms handling controlled unclassified information and registration in the System for Award Management.