Art of Novel Signals: Predicting and Forecasting with High Confidence
SBIR Opportunity Analysis
The Department of Defense, through DARPA, is seeking Direct to Phase II proposals under the SBIR program to develop a predictive geopolitical forecasting model utilizing temporal knowledge graphs and in-context learning. Performers must build a forecasting system that harvests and processes novel, multilingual open-source signals—such as radio, local news, and community audio—to bypass the data limitations of traditional AI models. The work requires establishing a radio data engine, fine-tuning automatic speech recognition for predominantly oral languages, implementing a synthetic data strategy, and integrating forecasting models to increase prediction timelines from days to weeks in data-sparse regions. The project targets strategic regions in Central and Southeast Asia, East and Northeast Africa, and South America, evaluating performance through sequential milestones across an 18-month base period and an optional six-month transition period. Proposals for this opportunity must be submitted by August 19, 2026, and the optional six-month period has an associated value of $500,000.