Generative and agentic AI: Production experience with LLMs for query understanding, conversational and agentic shopping, and content generation; retrieval-augmented generation; fine-tuning (LoRA, distillation); LLM serving and optimization with vLLM, TensorRT-LLM, or Triton; agent frameworks such as LangGraph; and rigorous LLM evaluation, guardrails, and cost management. Vector search and embeddings: Approximate nearest neighbor techniques (HNSW, IVF, product quantization) and libraries such as FAISS, ScaNN, or Milvus; dense and late-interaction retrieval models such as bi-encoders, two-tower models, and ColBERT; multilingual embedding models for cross-market search.