Prompt Engineering to AI Agents: Amazon Bedrock & Claude

Go from writing better prompts to building real AI agents — hands-on, in Python, on AWS Bedrock with Claude.

Most prompt engineering courses stop at tips and tricks. This one takes you all the way to production: by the end, you'll have built a RAG chatbot, a tool-calling assistant, a ReAct agent, a multi-step agent with self-correction and guardrails, and an evaluation pipeline that scores it all automatically.

What you'll learn, section by section:

- Prompting foundations — zero-shot vs. few-shot, chain-of-thought, role assignment, structured output, instruction clarity, multi-turn conversations, and multimodal prompts

- Advanced techniques — prompt chaining, routing, self-consistency, tree-of-thought, meta prompting, long-context strategies, and the top 10 anti-patterns to avoid

- Safety & quality — Bedrock Guardrails for content filtering, evaluation pipelines with LLM-as-judge, and how to measure quality with a number instead of a feeling

- RAG — retrieval-augmented generation with Bedrock Knowledge Bases, grounded on real documents

- Agents — tool calling and function calling, MCP servers, ReAct agents, and a full plan-and-execute agent with Reflexion-style self-correction and agentic guardrails

- Production skills — prompt caching to cut costs, DSPy, fine-tuning vs. prompting, AI ROI, and strategic prompting for decision makers

Every concept comes with a lab. You'll run real Python code against Claude on AWS Bedrock, break things on purpose, and learn why they fail — because that's how production skills actually stick.

Who this course is for:  developers, engineers, and technical professionals who want to build with LLMs — and decision makers who need to understand what's actually possible. Basic Python helps for the labs; no ML background required.

Requirements: an AWS account (labs cost a few cents in Bedrock usage) and basic Python familiarity.

Join me, and let's take you from prompting to shipping.