Z. Tschirhart, J.M. Rozmus
Sustainment,
United States
Keywords: large language model, vision language model, multimodal language model, retrieval augmented generation (RAG), defense manufacturing
Summary:
Sustainment is a venture-backed software and data science company that helps US-based manufacturers easily find and work with the critical suppliers they need to build and manage their supply chains. Our vision is to reimagine American manufacturing as a hyperconnected, secure, and resilient network of local and regional suppliers who can more easily connect, interact, and do business with the industry and government customers that rely on them. We support both DoD and commercial customers in pursuit of our vision and have extensive experience in defense supply chain operations. We are organized as a Public Benefit Corporation and have been awarded several R&D contracts and service multiple DoD customers. In 2024, large language models (LLM’s) and closely related vision language models (VLM’s) moved from being impressive chatbots that delighted consumers and fascinated computer scientists to business and engineering tools that can be specialized by teaching them a specific body of knowledge and setting them up to act as agents to assist computer users in a wide variety of tasks. This teaching is done by careful construction of queries, providing a collection of reference documents, and/or fine-tuning of the models. Large Language Models (LLMs) and Vision Language Models (VLMs) can significantly speed up the process of matching parts to new suppliers by leveraging their advanced capabilities in natural language understanding, image recognition, and data processing. Specifically, we are applying these AI models to interpretation of engineering drawings, researching the capabilities of manufacturers, and recommending the best suppliers to manufacture a specific part or assembly. This presentation will describe the design of these applications and the current level of performance.