Has built LLM evaluation using golden datasets, retrieval metrics (recall@k, MRR), faithfulness and groundedness checks, and LLM-as-judge, with tools such as Ragas, DeepEval or Promptfoo. · Has built agentic and multi-agent systems with LangGraph or a similar framework (CrewAI, AutoGen, Semantic Kernel), and understands state, checkpointing, conditional routing and.