AI agents are growing in their ability to reason and work autonomously over extended periods. As tasks become more complex and extensive, providing them with the right information is the main engineering challenge. The trend is moving from pre-loading context upfront to where agents dynamically retrieve data as necessary.
Redis is addressing context management with a context engine built on four pillars: on-demand context retrieval, constantly updated data, rapid retrieval, and an evolving memory layer. This involves creating materialized views with a semantic layer instead of allowing agents direct database access. A memory system extracts and compacts information asynchronously alongside the agent.
Simba Khadder, leading AI strategy at Redis and former co-founder of FeatureForm, joins Kevin Ball in this episode to discuss the prominence of context in agentic AI, differences between context engines and traditional RAG architectures, the role of materialized views in reliable data pipelines, improvements in memory systems through asynchronous methods, and how engineering practices must evolve as AI development speeds up.
Full Disclosure: This episode is sponsored by Redis.
Kevin Ball, also known as KBall, is VP of engineering at Mento and an engineering coach. He co-founded and was CTO of two companies, founded the San Diego JavaScript meetup, and organizes an AI discussion group through Latent Space.
Please click [here](http://softwareengineeringdaily.com/wp-content/uploads/2026/04/SED1923-Redis.txt) to see the transcript of this episode.