Anduril unleashes the power of RAG with enterprise search chatbot Alfred on AWS

In an age where data is the new currency, organizations are constantly seeking ways to harness the power of information to drive productivity and innovation. Anduril Industries, a defense technology company at the forefront of transforming military capabilities, has embraced this challenge head-on with the launch of Alfred, their internal enterprise search chatbot powered by cutting-edge Retrieval-Augmented Generation (RAG) architecture.

The RAG advantage: Combining the best of both worlds

RAG is an artificial intelligence (AI) framework that combines the power of retrieval-based systems with the generative capabilities of large language models (LLMs). By seamlessly integrating these two components, RAG offers an unparalleled ability to understand and respond to complex queries, using both the breadth of information in knowledge bases and the contextual reasoning abilities of LLMs. RAG employs a two-stage process. First, a retrieval component identifies relevant documents or passages from a vast knowledge base, narrowing down the search space. Then, a generative component, powered by an LLM, processes the retrieved information and generates a coherent and contextually appropriate response. This synergistic approach addresses the limitations of traditional search engines, which often struggle with ambiguous or complex queries, and those of LLMs, which can sometimes provide responses that lack factual grounding.

The data foundation: Key to generative AI applications

While the RAG architecture is a game-changer, its success hinges on the quality and breadth of the underlying data foundation. Recognizing this critical factor, Anduril has invested in curating and structuring its internal knowledge base, ensuring that Alfred has access to a comprehensive and up-to-date repository of information. By using Amazon Web Services (AWS) services, such as Amazon SageMaker, Amazon Kendra, and Amazon DynamoDB, on the secure AWS GovCloud (US) Regions, Anduril has built a robust and scalable data infrastructure that can support Alfred’s growing knowledge needs. With this data foundation, Alfred can not only provide accurate and relevant responses, it also possesses a strategic asset for future generative AI applications within the organization.

Security and compliance: Anduril’s unwavering commitment

As a defense technology company, Anduril operates in a highly regulated environment where data security and compliance are paramount. This is why the company chose AWS as its trusted partner, taking advantage of the robust security controls and compliance certifications offered by the AWS GovCloud (US) Regions. By hosting Alfred on the AWS GovCloud (US) Regions, Anduril ensures that sensitive information remains protected, with stringent controls on data access and user authentication. This approach not only meets the rigorous requirements of the defense industry but also aligns with Anduril’s mission to safeguard national security interests.

Customized solutions – Recognizing that different departments have unique needs, Alfred offers a selection of LLMs tailored to various use cases. Software, hardware, and business development teams can each use the model that best fits their specific requirements, giving them the tools to be their own boss.

Safe and compliant – Alfred is built to ensure that sensitive information remains protected, with stringent controls on data access and user authentication. In addition to stringent information security requirements for the defense industry, Controlled Unclassified Information (CUI) data requirements mean that data can’t leave Anduril systems, which historically ruled out commercial vendors for closed-source LLMs.

Adaptable and scalable – As LLMs and their uses evolve rapidly, Alfred’s design is intentionally open-ended, avoiding vendor lock-in and allowing the integration of the best and most advanced models as they become available for deployment in the AWS GovCloud (US) Regions. Alfred is more than just a tool—it’s a strategic asset that amplifies the capabilities of Anduril employees so they can perform at their best. It represents Anduril’s commitment to provide their employees with state-of-the-art technology, helping to make its internal systems advanced and forward-thinking as they work to build innovative defense solutions. This strategic approach ensures that Anduril remains at the forefront of enterprise search and knowledge management, enabling its employees to leverage the most cutting-edge technologies and stay ahead of the curve.

Unleashing employee productivity

The impact of Alfred on Anduril’s workforce has been nothing short of transformative. Employees across various departments, from software development to business operations, have reported significant time savings and increased productivity. At the time of its initial launch, Alfred had 300 unique users per day and 5,000 daily prompts. More than half of Anduril’s workforce has adopted Alfred. Employees are using Alfred for tasks like sifting through tens of thousands of pages of military standard documents (MIL STD) or summarizing and highlighting key points from complex Slack threads with hundreds of comments.

With Alfred’s ability to rapidly retrieve and synthesize information from disparate sources, employees can now effortlessly access the corporate knowledge graph without leaving their primary communication channels like Slack. This seamless integration has streamlined workflows, eliminating the need to sift through countless documents or Slack threads manually. Moreover, Alfred’s conversational interface allows employees to pose complex queries in natural language, providing them seamless access to the collective knowledge of the organization. In turn, employees can make informed decisions, collaborate more effectively, and drive innovation across the company.