This project aims to accelerate technology transfer by developing and demonstrating an advanced, artificial intelligence (AI)-driven platform for knowledge management and decision support called “Compass: The Environmental Knowledge Navigator.” Decades of investments by SERDP and ESTCP have generated a vast library of technical reports, guidance documents, and research findings addressing critical remediation challenges, including sites impacted with per- and polyfluoroalkyl substances (PFAS). Today, the scale and complexity of this knowledge base make it difficult for end users to efficiently locate, synthesize, and apply proven solutions.

The objective of this project is to bridge this gap by deploying Compass and enabling users to retrieve authoritative, context-specific answers from SERDP and ESTCP reports and other technology transfer products, such as videos, using natural language queries. As the capability of AI models continues to increase over time, it is also expected for Compass to provide even more transformative value by generating novel insights from SERDP and ESTCP data, rendering it a capable research assistant. In the initial phase, which is the focus of this project, information developed from more than 300 PFAS-focused SERDP and ESTCP projects will be utilized.

Compass will ingest existing SERDP and ESTCP PFAS media in all common formats (PDF, Word, tables, audio, and video), apply AI-based content understanding to normalize and chunk the material, and generate high-dimensional embeddings stored in a secure, vector database. A fine-tuned Large Language Model (LLM) operating in retrieval-augmented generation mode will ground each response in retrieved text segments, attaching inline citations that link directly to source pages. Authorized users will access the system via a secure web portal, natural-language query support, interactive citation display, and time-limited file downloads. This modular approach allows LLMs to be updated efficiently and enables files to be reindexed easily, ensuring the system's continued relevance in the evolving field of AI. It is also anticipated that future AI models will advance beyond retrieval to generate novel insights and act as effective SERDP and ESTCP PFAS research assistants. This is expected due to increased model intelligence associated with scaling test time compute and continual advancements in accomplishing long-horizon tasks.

By reducing information search and synthesis time from hours to seconds, Compass will enhance technology transfer and deliver immediate cost savings and efficiency gains, freeing technical staff to focus on higher-value tasks. Consistent, source-cited guidance will improve decision quality and uniformity across installations, maximizing SERDP and ESTCP's return on past research and development investments. Compass will also maximize the impact of prior research investments and break down institutional silos, fostering cross-installation collaboration, while accelerating the adoption of proven remediation technologies, ultimately bolstering operational capabilities and warfighter preparedness by mitigating the impacts of these chemicals. (Anticipated Project Completion - 2027)