Deborah A. Dahl

The paper proposes a memory-augmented, three-stage agentic pipeline that significantly reduces LLM hallucinations and improves operational efficiency by integrating semantic caching and advanced observability metrics.