🎥 Recorded live at the MLOps World | GenAI Summit 2025 — Austin, TX (October 9, 2025) Session Title: Memory and Memory Accessories: Building an Agent from Scratch Speaker: Robert Shelton, Applied AI Engineer, Redis Abstract: AI agents don’t have to be black boxes. In this live, hands-on demo, Robert Shelton from Redis shows how to build a fully functional, production-ready AI agent from scratch — no bulky frameworks required. Using only FastAPI, OpenAI, Redis, and Docket for async orchestration, Robert walks through how to create an agent capable of real, multi-turn conversations that leverage both short- and long-term memory. You’ll see how memory drives real-time reasoning, enables context retention, and powers retrieval and tool-calling — including live web search integration via Tavily. This practical session reveals how developers can build interpretable, maintainable, and performant agent systems using lightweight components and open-source tools. What you’ll learn: • How to design and deploy an AI agent from scratch using FastAPI, OpenAI, Redis, and Docket • The role of memory in enabling real-time reasoning and context retention • How to integrate short- and long-term memory for multi-turn interactions • Practical patterns for async task orchestration and scalable deployment on AWS • How retrieval and tool-calling (like Tavily search) enhance agent capabilities.

Memory & Memory Accessories Building an Agent from Scratch