Day 3 — RAG end-to-end

[PunithVT]

Day 3. Building on Day 1's Lambda — putting together a real RAG pipeline so we can both see how the AWS pieces and the retrieval pieces fit.

Topics to cover:

1. Embeddings 101 — what they are, picking a model (Titan, Cohere Embed, OpenAI)
2. Chunking strategies — fixed-size vs recursive vs semantic, and what breaks each
3. Vector storage on AWS — OpenSearch Serverless, pgvector on RDS, Bedrock Knowledge Bases
4. Retrieval + reranking — top-k, hybrid (BM25 + vector), cross-encoder rerank
5. Putting it together — a small Lambda + Bedrock + OpenSearch RAG chatbot

Plan: Chandana takes the AWS infra side, I take the retrieval / reranking side, we meet in the middle on the chatbot.