AI brokers are more and more able to reasoning and performing autonomous work over lengthy intervals. Nevertheless, as brokers tackle extra complicated, longer-horizon duties, retaining them equipped with the appropriate data turns into the core engineering problem. The trade is shifting away from pre-loading context upfront towards a mannequin the place brokers dynamically navigate and retrieve the info they want, after they want it.
Redis is approaching context administration utilizing a context engine, which is an structure constructed round 4 pillars: on-demand context retrieval, information that’s all the time present, quick retrieval, and a reminiscence layer that improves over time. In observe this implies constructing materialized views of knowledge with a semantic layer on high, fairly than giving brokers direct entry to manufacturing databases. A reminiscence system sits alongside this, extracting and compacting data asynchronously because the agent works.
Simba Khadder leads AI technique at Redis, and he beforehand co-founded the function retailer platform FeatureForm, which was acquired by Redis in 2025. On this episode, Simba joins Kevin Ball to debate why context has turn into the defining problem in agentic AI, how context engines differ from conventional RAG architectures, how materialized views underpin dependable agent information pipelines, how reminiscence methods can enhance by means of async extraction and compaction, and the way engineering groups must adapt their practices as AI-driven improvement accelerates.
Full Disclosure: This episode is sponsored by Redis.
Kevin Ball or KBall, is the vp of engineering at Mento and an impartial coach for engineers and engineering leaders. He co-founded and served as CTO for 2 corporations, based the San Diego JavaScript meetup, and organizes the AI inaction dialogue group by means of Latent Area.
Please click on right here to see the transcript of this episode.
