Objective

The objective of this topic is to apply Large Language Models (LLMs) and/or other Artificial Intelligence (AI) approaches to support and automate warfighter’s system’s integrations. This will pertain to problems with data unification and interoperability regardless of the target system, source system, or data format. It will focus on usage in tactical environments to assist and provide reliable performance, regardless of echelon level.

Description

In essence, data interoperability makes it feasible for data from numerous sources and formats to be integrated and leveraged together. An organization could derive the value from its data and overcome the major obstacles presented by distributed data assets by achieving data interoperability. The Department of Defense (DoD) recognizes efforts to improve interoperability, with three main topics of discussion: 1) a decrease in redundant and antiquated systems; 2) bridging the gap between DoD and mission partner standards; and 3) a more efficient and unified data flow.

The Army’s goal for the use and application of AI is to assist the national plan for leadership in AI-enhanced applications. Within the tactical environment, there exists numerous warfighting systems comprised of distinct databases and components. As more of these software’s develop and continue to be utilized, it is conceivable to employ LLMs and/or other AI approaches for handling interoperability that complies with government requirements and reduces harm to associated software from functional additions or modifications as software evolves to address interoperability challenges. The addition of new software to existing systems should be considered when improving interoperability.

The DoD is seeking to make ubiquitous data with one of these approaches which would result in data regardless of the sender’s system, recipients’ system, or data format, to be standardized across all systems in a tactical setting in a way that allows for this dialogue to function in both directions. Data may also pre-position based on AI-supported models.

For example, LLMs are a type of AI that are trained with deep learning models with vast and massive amounts of data. They could potentially bridge the gap in between systems. LLMs like ChatGPT can be leveraged to reach the efforts and goals of the DoD and Army by enhancing service interoperability and support for the warfighters in their missions. This strategy will function as a framework for software service optimization as needed by the warfighter’s circumstances.

Phase I

This topic is only accepting Phase I proposals for a cost up to $250,000 for a 6-month period of performance.

Research, document, and publish techniques, for training AI with respects to integration of traditional software systems. Identify an AI approach to integrate disparate software systems that includes but is not limited to API, data model and message mapping. Contractor will submit a list of metrics that are aligned to data interoperability goals that can be used to assess the performance of the proposed approach (Technology Readiness 2).

Phase II

Provide a concept demonstration to address interoperability issues with traditional software systems. This concept demonstrator should show that two or more traditional systems can be made more interoperable by using this technique, minimizing the need for software customization. The offeror should show how the approach measured up against the aforementioned metrics reflects the improvements created relating to interoperability. Metrics outlined in Phase I will be updated as required (TRL 5).

Phase III

This technology could easily be adapted with a different training data set in order to increase interoperability of commercial systems, including:

·       Finance: Requires real-time analytics and risk assessments from multiple systems

·       Customer support: LLM agents can help integrate multiple data points and analytics from various web browsers/systems

·       Cybersecurity: For various markets, cybersecurity professionals would benefit from NLP querying from a myriad of like databases and cyber intrusions

·       Healthcare: Similar to finance, doctors and pharmaceuticals would benefit from those efficiencies

Submission Information

For more information, and to submit your full proposal package, visit the DSIP Portal.

SBIR|STTR Help Desk: usarmy.sbirsttr@army.mil