An AI Engineer at a retail company is developing a customer support AI agent that...

The selected option specifically C states “Implement a hybrid memory system with vector-based search and key-value storage to retrieve relevant past interactions.”, which matches the operational requirement rather than a superficial wording match. Hybrid memory lets the agent combine fast key-value facts with semantic vector recall. Expanding the context window is the blunt and expensive alternative. The architecture implied by Option C is the one that survives real workloads: separate responsibilities, explicit contracts, and measurable runtime behavior. In NVIDIA terms, agentic workflows need explicit state management; external memory complements the LLM context window while fine-tuning encodes stable behaviors into model policy. The correct implementation surface is external state stores combined with model adaptation when repeated behavior should become part of the policy. That is why the other options are traps: a single flat store cannot serve both low-latency conversational state and durable semantic recall equally well. This choice gives engineering teams the knobs they need for continuous tuning after deployment.