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

The contract involves the development, configuration, and optimization of a Retrieval-Augmented Generation (RAG) large language model leveraging Azure OpenAI technology. The primary goal is to enable the system to ingest data from multiple sources, retrieve relevant information from both vector and graph databases, and generate outputs that include properly cited and traceable sources. This effort supports enhanced information synthesis and retrieval capabilities within the Department of Defense. The work is designated as a subcontract under the NAICS code 511210, focused on software publishers. The place of performance is Fort Belvoir, with the Department of Defense’s Disa PL8 office overseeing the contract. The project emphasizes advanced AI model tuning to improve retrieval accuracy and output reliability, facilitating better decision-making through transparent and verifiable AI-generated content.