Index · Every exercise and lab, by track and level

[akash-coded]

Fifty-plus practice threads is a lot to scroll. This is the map.

Every thread below is tagged, so you can also filter the list directly:
🟢 foundational · 🟡 intermediate · 🔴 advanced · 💻 code · ✏️ no-code

🟢 foundational · 🟡 intermediate · 🔴 advanced

---

🧭 Foundations

_{Agent vs workflow, model intuition, scoping}

• 🟢 LLM Intuition Bank · Build model intuition without any maths · _exercise

📋 Product & PDLC

_{Artefacts, gates, and the decisions before code}

• 🟢 Agentic PDLC 1 · Triage bench: what kind of statement is this? · _exercise
• 🟢 Agentic PDLC 2 · Ranking, rungs, and the cost of delay · _exercise
• 🟢 Agentic PDLC 4 · Four calculations, in order · _exercise
• 🟡 Production Recap 4 · The lifecycle, the gates and the ambition ladder · _exercise
• 🟡 Design on Paper · Control and execution planes, from PRD to HLD to LLD · _lab
• 🟡 Agentic PDLC 3 · Case file: a cancelled flight, and seven questions · _exercise
• 🔴 Architecture Decision Lab · Designing a production agentic workflow before writing code · _lab

🪨 Amazon Bedrock

_{Converse API, tokens, tool use, guardrails}

• 🟢 Bedrock Model Selection Lab · Choosing a model you can defend on four dimensions · _lab
• 🟢 The Verbosity Tax · Why two identical questions can cost 7× different · _exercise
• 🟡 Bedrock Converse Lab · Build a disruption desk and reason about the agent loop · _lab

🔁 Agent Loop

_{Dispatch, history, termination}

• 🟡 Bedrock Converse Lab · Build a disruption desk and reason about the agent loop · _lab

🔧 Tools

_{Schemas, contracts, honest failure}

• 🟡 LangChain Agents 2 · Tools, routing, and the failure you have to handle · _exercise
• 🟡 LangChain Applied · Giving an agent hands without giving it the keys · _exercise
• 🟡 Strands Build 2 · The toolsmith: schemas the model calls correctly · _exercise

🧵 Frameworks

_{Strands, LangChain, LangGraph, LiteLLM}

• 🟢 LangChain Agents 1 · From a bare model call to an agent with hands · _exercise
• 🟢 LangChain Coding 1 · Agents and pipes: composing runnables that earn their place · _exercise
• 🟢 LLM Memory 1 · Build the notepad: statelessness, RAIL and session isolation · _exercise
• 🟢 LiteLLM 1 · One interface across providers, and what it hides · _exercise
• 🟢 Strands Build 1 · Give the agent a job, then stop it guessing · _exercise
• 🟡 LangChain Agents 2 · Tools, routing, and the failure you have to handle · _exercise
• 🟡 LangChain Applied · A triage pipeline that routes before it reasons · _exercise
• 🟡 LangChain Applied · Giving an agent hands without giving it the keys · _exercise
• 🟡 LangChain Coding 2 · Composition and control flow you can test · _exercise
• 🟡 LiteLLM 2 · Fallbacks, retries and routing that does not degrade silently · _exercise
• 🟡 Strands Build 2 · The toolsmith: schemas the model calls correctly · _exercise
• 🟡 Strands Build 3 · Open sandbox: build your own agent · _exercise
• 🔴 LangChain Intermediate · FreightDesk: one incident, nine fixes · _lab

🧠 Memory

_{Statelessness, buffers, retention}

• 🟢 LLM Memory 1 · Build the notepad: statelessness, RAIL and session isolation · _exercise
• 🟡 LLM Memory 2 · Ship it without forgetting: instance amnesia and the replay tax · _exercise

📚 Retrieval & RAG

_{Chunking, hybrid search, reranking, citations}

• 🟢 RAG 1 · Why your FAQ bot invents answers, and what naive retrieval fixes · _exercise
• 🟡 RAG 2 · Query rewriting, reranking and hybrid search — measured, not assumed · _exercise
• 🔴 Capstone Take-Home · Strands, A2A, A2UI, AgentCore, LiteLLM and RAG in one build · _capstone

🕸️ Multi-Agent

_{Delegation, critique, orchestration patterns}

• 🟢 Agentic Workflow Patterns 1 · Who controls the control flow, you or the model? · _exercise
• 🟡 Agentic Workflow Patterns 2 · The three dials and the bill they produce · _exercise
• 🟡 Agentic Workflow Patterns 3 · Delegation and critique, and what each costs · _exercise
• 🔴 AgentCore Lab B · VoltDesk: code interpreter, multi-agent, and identity · _lab
• 🔴 Architecture Decision Lab · Designing a production agentic workflow before writing code · _lab
• 🔴 Agentic Patterns Compound Lab · Choosing and wiring seven patterns · _lab
• 🔴 A2A Capstone · Aurora Grid: an outage desk built from independent agents · _capstone
• 🔴 Capstone · Switchboard: an LOB-pluggable incident triage desk · _capstone

⚙️ AgentCore

_{Runtime, memory, identity, gateway}

• 🟡 AgentCore Lab A · ParcelPilot: runtime, memory, and a bounded definition of done · _lab
• 🔴 AgentCore Lab B · VoltDesk: code interpreter, multi-agent, and identity · _lab
• 🔴 Capstone Take-Home · Strands, A2A, A2UI, AgentCore, LiteLLM and RAG in one build · _capstone

🔌 Interop (A2A / A2UI)

_{Agent cards, handshakes, agent-driven UI}

• 🟢 A2A 1 · Read the agent card before you trust it · _exercise
• 🟡 A2A 2 · The two-agent handshake, end to end · _exercise
• 🔴 Capstone Take-Home · Strands, A2A, A2UI, AgentCore, LiteLLM and RAG in one build · _capstone
• 🔴 A2A Capstone · Aurora Grid: an outage desk built from independent agents · _capstone

🔬 Evaluation & QA

_{Golden sets, judges, gates}

• 🟡 Production Recap 5 · Proving it: golden sets, judge bias and trajectory eval · _exercise
• 🔴 Capstone · Ship RailReserve: a full agent build in a new domain · _capstone
• 🔴 QA Lab · Test TravelMind end to end and gate the release · _lab

🚀 Production

_{Deploy, fail over, roll back, observe}

• 🟢 Production Recap 1 · The eight-layer map and where each decision belongs · _exercise
• 🟡 LiteLLM 2 · Fallbacks, retries and routing that does not degrade silently · _exercise
• 🟡 Production Recap 4 · The lifecycle, the gates and the ambition ladder · _exercise
• 🔴 Capstone · Ship RailReserve: a full agent build in a new domain · _capstone
• 🔴 Capstone · Switchboard: an LOB-pluggable incident triage desk · _capstone

---

Where to start

If you are…	Start with
New to all of this	LLM Intuition Bank, then Agentic PDLC 1 — neither needs an AWS account
An engineer	Bedrock Converse Lab → Strands Build 1 → Agentic Workflow Patterns 1
Working on RAG	RAG 1 → RAG 2 → QA Lab
A PM, BA or architect	The four Agentic PDLC threads, then Design on Paper
Ready for a capstone	RailReserve · Aurora Grid · Switchboard

Prefer something graded?

The L.A.B. Simulator is the auto-graded counterpart to these threads: bite-sized labs with public and hidden checks, a CLI runner, and a Break phase that tries to break your working solution. Runs offline, no AWS account.

python labs/runner/labctl.py next

---

Posting your work? Reply in the thread itself, not here. Stuck? Q&A. Something missing? Ideas.