SRI Faculty Luncheon & Workshop: Building Space Mission Control Agents with Amazon Bedrock

Workshop Overview

Welcome to the Workshop on Bedrock Knowledge Bases and Agents for Developing Advanced Space Mission Control Chatbots, co-sponsored by the Space Research Institute and Amazon Web Services.

In this workshop, you will learn how to harness the power of Amazon Bedrock Knowledge Base, agents, and guardrails for space mission operations.  While this workshop focuses on space mission operations, you can envision extending this solution to other areas of aerospace and defense, such as satellite operations, ground station management, or space exploration planning.  The target audience are those who have an interest in learning about generative AI, knowledge bases, and the use of AI agents. You should have a basic understanding of generative AI solutions and be comfortable navigating the AWS console to deploy services. 

You will start by building a Knowledge Base that contains information about Leo, Amazon's satellite broadband network. This Knowledge Base will serve as a centralized repository, allowing users to ask questions about the service and receive accurate, up-to-date responses with reference to source material.

Next, you will develop an intelligent agent capable of understanding natural language queries to perform tasks, such as retrieving information from a spacecraft database and calculating orbital parameters . One can imagine extending the agent's capabilities to performing tasks on behalf of mission controllers, like analyzing spacecraft telemetry data and determining system health status based on current readings. Agents can also initiate diagnostic procedures, retrieve mission parameters, or interact with spacecraft systems to streamline mission operations.

Mission security and safety should always be a top concern, and the workshop will wrap-up with the creation, implementation, and testing of Amazon Bedrock Guardrails. You will enable safeguards to detect and filter harmful content by applying content filters and denied topics along with enabling prompt injection attack filters. You will then input prompts to test the guardrails and see confirmation of detected activity outside of policy.

Additionally, you can explore integrating this solution with existing mission control systems or other aerospace applications, creating a unified and intelligent platform for space operations. The potential for leveraging generative AI in streamlining complex space missions and enhancing operational efficiency is expansive, and this workshop serves as a stepping stone towards unlocking that potential.