Contract Award LAS 2024

This research will directly support LAS’s grand challenge problem to “generate tailored daily reports for knowledge workers that capture information relevant to their individual objectives and interests.” This challenge is the multi-year focus of LAS’s annual Summer Conference on Applied Data Science (SCADS). Dr. Chris Argenta, founder of Rockfish Research, recently participated in SCADS and created a TLDR prototype called “Relevance” to demonstrate using a knowledge graph to inform a large language model (LLM) and drive analytics to tailor a news-like data stream and generate simple but concise TLDR. While LLMs are both powerful and popular today, they are not known for sticking to the facts (i.e., hallucinations) and often generate bloated responses. Relevance tailors unstructured text content down to an accurate selection and abstract summarization of material that is most relevant to user’s stated interests. It accomplishes this using a knowledge graph to drive each step of the workflow, and this integration helps overcomes many of the technical issues that AI components may experience individually.

In our 2024 project, Rockfish Research will collaborate with government, academic, and industry researchers to further develop these concepts into a complete, flexible, and open AI-driven workflow system that supports plug-in analytics, provides interfaces for diverse experimentation, and allows researchers to rigorously evaluate their work within an end-to-end TLDR. This framework will empower SCADS participants (and other researchers) to collaborate efficiently and effectively towards solving the full scope of this multi-year grand challenge problem.

We are very excited to be working with the folks at LAS. Please plan to come see us and other intelligence community-focused researchers at the LAS Research Symposium at NC State in December 2024.