Unlocking the Data Layer for Agentic AI with Simba Khadder

As AI agents evolve from short, supervised tasks to autonomous, hour-long problem solvers, the defining challenge in software engineering is no longer model intelligence—it's context. Simba Khadder, AI Strategy Lead at Redis and co-founder of Featureform, argues that traditional RAG (Retrieval-Augmented Generation) is obsolete for complex, long-horizon agent workflows. Instead, the future lies in 'context engines'—architectures built on four pillars: on-demand context retrieval, always-up-to-date data, fast access, and memory systems that improve over time. These engines rely on materialized views of data, not direct access to production databases, with Redis serving as a semantic layer that transforms raw data into agent-friendly, human-readable structures. Crucially, memory isn't just storage—it's an async, LLM-powered ETL process that extracts, compacts, and refines agent behavior over time. The real shift isn't just technical: engineering teams must now prioritize system design, behavior testing, and architecture reviews over line-by-line code scrutiny, as agents generate vast amounts of code quickly. The most powerful teams aren't those with the best tools, but those who master the feedback loops between spec, behavior, and memory—turning context into a moat that separates leaders from followers in the agentic era.