• Building AI search agents- including ReAct, planning, and multi-agent architectures via custom implementation or frameworks like LangGraph, Dify, or CrewAI. • Building end-to-end RAG pipelines from ingestion, chunking, embeddings, and hybrid vector search, ideally using Opensearch. • Operating and monitoring vector/hybrid indexes (e.g. OpenSearch) in production environments. • Implement grounding and citation to link generated answers back to their exact source passages. • Automate evaluation using synthetic QA, retrieval-hit-rate tracking, and model-critique loops to continuously measure accuracy and detect drift. • Orchestrating external tools or knowledge bases and monitoring latency and cost at production scale.