Course Companion

The course companion

Sixteen modules, five tracks, one rule: every abstraction is preceded by the thing it abstracts.

The course lives in the repository, under modules/: slides, notebooks, exercises, workbooks and solutions, reviewed and versioned. These pages are the companion: one page per module with its reading order, the tutorial lesson that frames it, the workbench tool that exercises the same decision, and where its errors and questions are collected. Read the module README first, always; this page tells you what to read around it.

Where to start: START-HERE in the repository picks a path for you. Not sure? Module 00 needs no AWS account and no code.

The five tracks

Track	Modules
Think	00–01
Invoke	02–04
Build	05–09
Ground	10
Ship	11–14
Product · parallel	15

You write the agent loop by hand before meeting a framework (05 → 06), build retrieval by hand before a managed knowledge base (10), and test before you deploy (13 → 14).

All 16 modules

#	Track	Module	Time	Read first
00	Think	Agentic Foundations Decide what deserves to be an agent before you build one.	4–5 hours	P0 · Frame
01	Think	LLM Intuition and the AWS Bridge Build a working mental model of LLMs, then map it onto AWS.	3–4 hours	What is the agentic PDLC?
02	Invoke	Amazon Bedrock Essentials Your first real calls: Converse API, tokens, tool use, knowledge bases, guardrails.	6–8 hours	Why the AI bill is 4× the estimate
03	Invoke	Amazon Bedrock Agents Console to code: action groups, orchestration, and controlling behaviour.	6–7 hours	P1 · Design & Spec
04	Invoke	Agent Builder, Knowledge Bases and Guardrails The low-code surface, and how to wire knowledge and safety into it.	4–5 hours	Guardrails that hold
05	Build	The Agent Loop: No Framework to Strands Write the loop by hand, feel the pain, then let Strands remove it.	5–6 hours	P2 · Build & Prove
06	Build	Strands Foundations: Tools, Memory and MCP Agents with hands, memory, and a standard way to reach the outside world.	5–6 hours	For engineers
07	Build	Multi-Agent Patterns with Strands Swarm, graph, delegation, critique — and when each one is wrong.	6–7 hours	For solution architects
08	Build	LangChain and LangGraph The other ecosystem — and an honest side-by-side with Strands.	7–8 hours	One lifecycle for every method
09	Build	LLM Memory Mechanics What models forget, why, and what you can do about it.	3–4 hours	Why the AI bill is 4× the estimate
10	Ground	RAG, OpenSearch and LiteLLM Retrieval done properly — chunking, hybrid search, reranking, and an evaluation gate.	8–10 hours	Prove the agent meets its bar
11	Ship	Amazon Bedrock AgentCore Runtime, memory, identity, gateway, observability — agents as deployed services.	8–10 hours	P3 · Run & Learn
12	Ship	A2A and A2UI: Agent Interoperability Agents talking to agents, and agents talking to users.	4–5 hours	P1 · Design & Spec
13	Ship	Agentic QA and Evaluation How you prove an agent works — and block the ones that do not.	5–6 hours	For QA
14	Ship	End-to-End Production Pipeline The capstone: build, validate, deploy, fail over, and gate a release.	8–10 hours	Shadow mode and cut-over
15	Product · parallel	Agentic Product Lifecycle For the people who decide what gets built, and have to defend it.	4–5 hours	For product managers

Pick a path

Nobody needs all sixteen in the same order. Each path is a subset, in a sequence, with a finish line:

Path	For	Time	Finish line
🚀 Weekend Sprint	"Show me it works, this weekend"	~12 h	A working Bedrock agent with tools
🛠️ Agent Engineer	Building and shipping agents	~70 h	A deployed, evaluated, gated agent
🏛️ Solutions Architect	Designing and reviewing agent systems	~45 h	A defensible architecture with a cost model
📋 Product Manager	Deciding what to build and defending it	~25 h	A PRD and a gate review you can run
🔬 RAG Specialist	Retrieval quality as the job	~35 h	A measured, gated RAG pipeline

1. Attempt exercises closed-book. Solutions are written to be read after you have a wrong answer to compare against. Reading them first converts an hour of learning into five minutes of nodding.
2. Fill in the workbooks. The .xlsx activities are where a decision gets written down and costed. Skipping them is how you end up unable to defend the design later.
3. Do not skip Module 05. Writing the agent loop by hand is what stops every later framework from being magic.
4. Watch the bill. Set a budget alarm before Module 02. See cost controls.

Prove it, then teach it

• The labs companion: auto-graded labs that try to break your solution, and which module teaches each one.
• The cohort kit: ninety-minute sessions that turn the tutorial and this course into a programme for a team.
• Study plans: the same material fitted to a week, a weekend, four weeks of evenings or twelve weeks of reading group.

---

_{✏️ This page is generated from modules/README.md, the module READMEs and docs/learning-paths/README.md: an edit made here is replaced at the next sync.}