One of the primary difficulties in designing AI agents is the stateless nature of large language models, along with their limited context windows. This creates a need for precise engineering to ensure consistent and reliable interactions across sequential LLM sessions. Successful agents require efficient systems for storing and accessing short-term conversations, summaries, and long-term information.
Redis, an open-source, in-memory data store, is often employed for high-performance caching, analytics, and message brokering. Recent developments have expanded its functionality to include vector search and semantic caching, further establishing it as a key component in the agentic application stack.
Andrew Brookins, a Principal Applied AI Engineer at Redis, joins Sean Falconer on the show to explore the challenges of building AI agents, the role of memory in these systems, the difference between hybrid search and vector-only search, the idea of world models, and more.
Note: This episode is sponsored by Redis.
Sean has experience as an academic, startup founder, and Googler, with published work in fields ranging from AI to quantum computing. He is currently an AI Entrepreneur in Residence at Confluent, focusing on AI strategy and thought leadership. Connect with Sean on LinkedIn.
Access the episode transcript [here](http://softwareengineeringdaily.com/wp-content/uploads/2025/08/SED1868-Andrew-Brookins.txt).
For sponsorship inquiries: [[email protected]](https://cms.megaphone.fm/organizations/619b680e-d813-11ea-9750-e7ecac406436/podcasts/e60108fe-e328-11ea-b2b6-1348249f900a/episodes/3913370a-bc44-11ed-ace6-8b1f96d82bba/[email protected]).