Fusion of Abstract Learning and Context-Optimized Neural-methods (FALCON)

This federal solicitation (DPA26BZ04-DV016) details a Direct to Phase II (DP2) SBIR/STTR opportunity titled "Fusion of Abstract Learning and Context-Optimized Neural-methods" (FALCON). The program's objective is to combine advanced machine learning (ML) methods for structured data with large language models (LLMs) to enable powerful, interactive, and computationally efficient statistical analysis of large-scale datasets, such as those found in enterprise or battlefield environments. Key technical requirements include researching suitable ML methods, developing an architecture that integrates ML with LLMs, mitigating model hallucinations, demonstrating reproducible analytical traces, and scaling the solution to enterprise-level data. The projected Cybersecurity Maturity Model Certification (CMMC) requirement is Level 2 (Self). Because this is a Direct to Phase II solicitation, proposers must bypass Phase I by proving prior comprehensive research experience and technical feasibility from the last three years in their proposal. Phase II expectations include developing the software, demonstrating initial functionality within six months, scaling to enterprise-level data with at least two distinct datasets by month 12, and creating a parallel commercialization and go-to-market plan. The performance period includes a 12-month base period with specific milestone deliverables and a 6-month option period ending at month 18. Successful technology must have dual-use applications, offering superior performance over current state-of-the-art methods for both civilian commercial markets and military strategic or tactical operations.