Retrieval-Augmented Generation (RAG) Large Language Model Development and Tuning

The contract entails the development, configuration, and optimization of a Retrieval-Augmented Generation (RAG) powered large language model using Azure OpenAI services to support advanced data processing capabilities. The system must ingest and process information from multiple data sources, leveraging both vector and graph databases to retrieve contextually relevant information, ensuring the generated outputs are grounded in accurate, traceable sources. A critical requirement is the production of cited responses with full provenance tracking, allowing users to verify the origin of every generated statement. The solution will be tailored for high reliability and performance in a secure defense environment, with precise tuning to meet operational needs for accuracy, latency, and scalability. The work is a subcontract under the Department of Defense’s Disa PL8, with performance located at Fort Belvoir, Virginia, ZIP 22060. The project falls under NAICS code 511210, indicating it is focused on software publishing and related services, specifically custom software development for AI and data-driven applications. The contract was posted on May 13, 2026, and is intended to deliver a production-grade RAG system capable of handling complex, multi-source intelligence data while maintaining strict traceability and auditability of all generated content. The implementation will integrate tightly with existing DoD infrastructure and adhere to stringent security and compliance standards associated with federal defense contracts.