AI-Enabled “Just-in-time” Knowledge Delivery for Naval Maintainers

This effort is aimed at addressing the critical operational challenge of maintaining increasingly diverse and technologically complex arrays of systems with generalist knowledge. The solution requested in this SBIR topic is a just-in-time knowledge delivery system for Fleet maintainers that tells them what they need to know, when they need to know it. It would provide detailed procedures and technical knowledge for diagnostics, repair and maintenance procedures on demand, both ashore and at sea, where there will be limited access to remote data stores. The system will employ generative AI to analyze authoritative technical documentation and historical maintenance records, at both the Fleet and individual asset levels, and generate summarized technical guidance initiated by free-text queries or guided diagnostic dialogues. Post-task, the system will prompt the technician for repair feedback, synthesizing their insights to update the database with both general maintenance guidance and individual asset histories. Problems with current knowledge delivery are particularly acute when new systems and technology are introduced to the Fleet, such as with the introduction of unmanned systems (UxS), and when locally developed knowledge needs to be propagated across the larger maintenance community. Operational readiness is currently bottlenecked by an inefficient maintenance paradigm: generalist maintenance technicians constantly face unfamiliar equipment without access to historical repair data, while frequent duty rotations continually erase localized (tribal) knowledge. This over-reliance on a transient pool of specialists causes extended system downtime and degrades operational readiness. These problems are particularly acute when new systems and technology are introduced to the Fleet. While new advances in Large Language Models (LLMs) and generative AI present an opportunity to create sophisticated human-machine teaming tools for providing real-time interactive support, state-of-the-art predictive maintenance systems are not attuned to the unique challenges of the Navy’s maintenance environment. Effective solutions will need to integrate diverse approaches and overcome existing hard science problems to arrive at the desired solution. Current approaches in this domain excel at forecasting component failures and recommending procedures; however, they lack the interactive step-by-step guidance required to conduct the procedures or for dynamic on-the-job technical support to the maintainers. Existing maintenance knowledge tools operate as one-way information pipelines that fail to capture critical institutional knowledge and platform-specific maintenance histories, which prevents the continuous, enterprise-wide learning loop necessary for resilient maintenance operations. Solutions must be able to ingest, synthesize, and appropriately structure a wide variety of data formats (e.g., technical manuals, structured databases, handwritten logs) while also integrating newly produced recurrent maintenance reports. This data should serve as a dynamically updated foundational knowledge base that is contextualized by the physical and operational situation unique to each repair, providing customized guidance to the technician. The solution should adapt its guidance to the user's specific skill level and experience to ensure interpretability across levels of expertise. Lastly, the desired Maintenance Associate solution must be capable of operating on local hardware in low-bandwidth and disconnected environments, operating on local storage and syncing fleetwide as connectivity allows. This SBIR topic is seeking proposals that would contribute to the development of an AI-Enabled Maintenance Associate. It is expected that the initial capability would be focused on a specific technical domain and a representative set of equipment that would serve as a prototype that can be scaled up to apply to multiple domains and an array of deployed systems. One target domain of immediate interest is the maintenance and repair of unmanned systems. However, any equipment and application domain that is found within Naval systems is of interest. The envisioned system workflow will ingest and distill technical documentation into interactive, multi-media or hard-copy repair procedures—complete with parts lists, safety warnings, and annotated tribal knowledge—delivered through a user-friendly Human-Machine Interface (HMI). The system will capture post-repair outcomes to continuously update the foundational knowledge base. Proposed solutions may address individual components of the Maintenance Associate system or work to develop a complete system. Components of the Maintenance Associate System that may be addressed for this SBIR topic should include: • A database management system for repair service/technical manuals and logs for distillation into maintenance procedures. • An HMI for technician requesting support that delivers summary procedure, to include tools, parts lists, cautions, and warnings; and provides results that are appropriately annotated localized (tribal) knowledge (best practices, tips, and comments from other users). • System products to end-users delivered as: multi-media content and/or hard copy outputs for technicians to access when performing procedures. • An incorporated mechanism to capture results/maintenance outcomes with a data integration mechanism that feeds into the knowledge base for future reference. Ideally, this transition effort will specifically focus on supporting the complex maintenance, diagnostics, and repair requirements of UxS hardware and platforms (e.g., Unmanned Surface Vessels, Unmanned Underwater Vehicles, Unmanned Ground Systems, and Unmanned Aerial Systems), although use cases for other Naval Applications and alternative transition customers would be acceptable. Unmanned systems and commands are nascent, and subject to wholesale platform change in short order. Operators and maintainers rely on vendor provided maintenance doctrine as they are developing historical context. Grounding for the Associate must include these documents which may be CUI or SECRET, performers must be able to operate at required classification levels. Reported technical evaluations, capability demonstrations, and success metrics will focus on the accuracy/interpretability of maintenance guidance and quality of the human-machine interaction as assessed by real-world end users and the generalizability of the technical approach across platforms and domains. Offerors are reminded that SBIRs are expected to sufficiently develop one or more capabilities that will receive endorsement from an acquisition/development organization and who will provide matching funds for transition/acquisition during Phase II of the SBIR. Prospective agencies for potential matching funds should be identified in the Phase I proposal along with the description of the baseline capability to be delivered that the proposer believes that acquisition agency would be interested in funding during Phase II. The proposer will need to: 1) Develop an AI-driven, maintenance task context-aware guidance prototype that functions as an on-demand technical expert and centralized intelligence hub for maintenance technicians, providing interactive support for complex diagnostic and repair tasks. 2) Incorporate a mechanism to ingest, process, and index a wide array of reference materials, including technical manuals, vendor publications, Maintenance Required Cards (MRCs), after-action reports, and historical service logs from various databases. 3) Demonstrate a workflow for capturing new experiential knowledge, observations, and "lessons learned" from maintainers post-task to ensure the system continuously learns and shares updated information across the fleet. 4) Be designed to operate effectively on simple hardware (e.g., tablets, laptops) in environments with limited or no connectivity by using a local knowledge store, with the capability to synchronize with a central cloud database when a connection is available. 5) Provide an adaptive human-machine interface that guides a user through troubleshooting and repair via natural language, tailors instructions to the technician’s skill level, and recognizes the specific system variant in use. Work produced in Phase II may become classified. Note: The prospective contractor(s) must be U.S. owned and operated with no foreign influence as defined by 32 U.S.C. § 2004.20 et seq., National Industrial Security Program Executive Agent and Operating Manual, unless acceptable mitigating procedures can and have been implemented and approved by the Defense Counterintelligence and Security Agency (DCSA) formerly Defense Security Service (DSS). The selected contractor and/or subcontractor must be able to acquire and maintain a secret level facility and Personnel Security Clearances. This will allow contractor personnel to perform on advanced phases of this project as set forth by DCSA and ONR in order to gain access to classified information pertaining to the national defense of the United States and its allies; this will be an inherent requirement. The selected company will be required to safeguard classified material during the advanced phases of this contract IAW the National Industrial Security Program Operating Manual (NISPOM), which can be found at Title 32, Part 2004.20 of the Code of Federal Regulations.