9:20 Thaddeus Worsnop, Solutions Architect, AWS

Supercharging LLMs: Delivering Value with RAG and Bedrock Agents

Large language models (LLMs) have demonstrated remarkable abilities in natural language understanding and generation. However, LLMs are limited by the knowledge contained in their static training data. Retrieval-augmented generation (RAG) is a technique that enables LLMs to access and incorporate domain specific data or proprietary information securely from various knowledge sources, significantly extending their capabilities. RAG retrieves relevant passages from a knowledge base and provides them as additional context to the LLM, allowing it to generate outputs grounded in current, factual information.

Amazon Bedrock agents further enhance RAG by orchestrating multistep workflows involving foundation models, knowledge bases, and API calls. Agents can break down complex user queries, retrieve pertinent data via RAG, reason over the retrieved information, and generate comprehensive responses - all without manual engineering. The agent traces its reasoning steps, enabling developers to refine its behavior.

This session will explore how RAG and Bedrock agents can be leveraged to supercharge LLMs for knowledge-intensive applications. We will examine the key components of a RAG architecture, best practices for constructing knowledge bases, and methods for optimizing retrieval. Example applications in question answering, task completion, and open-ended dialogue will demonstrate the power of retrieval augmentation. Finally, we will discuss practical considerations and future directions for RAG and agent-based systems.

By harnessing external knowledge through RAG and Bedrock agents, we can push the boundaries of what LLMs can achieve, unlocking more capable, reliable, and scalable language technologies. Attendees will leave with an understanding of core RAG concepts and concrete ideas for implementing retrieval augmentation in their own project.