Building AI applications on AWS — one project at a time.
Building AI applications on AWS — one project at a time.
In Part 7, we built a multi-agent support system with Strands Agents — a supervisor that routes customer requests to three specialist agents, with human-in-the-loop safety for sensitive actions. It works on your machine. Now how do you run it...
In Part 6, we built a support assistant with guardrails, intent detection, and conversation history. It worked — but the routing was Python keyword matching. “Charge” goes to billing. “Password” goes to account. Simple, brittle, and wrong the moment a...
In Part 5, we optimized RAG search with hybrid retrieval, reranking, and evaluation metrics. The search is smart. But a smart search engine isn’t a support assistant. It doesn’t know who it is, what it should refuse to answer, or...
In Part 4, we built a RAG knowledge base with vector stores, embeddings, and document processing. It works — ask a question, get an answer grounded in your documents. But “works” and “works well” are different things.
In Part 3, we built a pipeline that processes text, images, audio, and surveys. All of it fed into a single Bedrock call for analysis. But the knowledge behind that analysis was baked into the prompt — no persistent memory,...