Deploy an AI Project on LocalStack

• Context: what are we going to build (and why)?
• Where each piece runs (and why)
• The session plan
• Key vocabulary

• Phase 0 — Setup: Docker and LocalStack running (25-40 min)
• Step 1: Docker — from zero to engine running
• Step 2: Create the project and launch LocalStack
• Step 3: Install the CLI tools (there are TWO, and you install both)
• Step 4: Fake credentials
• Step 5: Verify the connection
• Step 6: Stopping and starting (learn it now)
• ✅ Phase 0 Checkpoint

• 1.1 Create the sample documents
• 1.2 Create the two buckets
• 1.3 Upload the documents
• What you just did, in one picture
• ✅ Phase 1 Checkpoint

• Phase 2 — The `doc-summarizer` Lambda (50 min)
• 2.1 Design: one function, two doors
• 2.2 The code
• 2.3 Packaging
• 2.4 IAM role
• 2.5 Choose the model and create the function
• 2.6 First invocation (no API Gateway yet)
• 2.7 Tests: the logic is tested without infrastructure (10 min)
• ✅ Phase 2 checkpoint

• Phase 3 — API Gateway: `POST /summarize` (20 min)
• 3.1 Create the API and connect it
• 3.2 Test it over HTTP

• Phase 4 — Event-driven: S3 triggers the Lambda (25 min)
• 4.1 Give S3 permission to invoke the Lambda
• 4.2 Configure the notification on the bucket
• 4.3 The moment of truth
• ✅ Phase 4 Checkpoint
• `recreate.sh`: your insurance against restarts (community is ephemeral)
• Midpoint — from pipeline to full AI application

• Phase 5 — RAG indexing: the document becomes searchable (30 min)
• 5.1 Extend the IAM permission
• 5.2 Add indexing to `service.py`
• 5.3 Repackage and update the Lambda
• 5.4 Build the index
• ✅ Phase 5 Checkpoint

• Phase 6 — The agent: `POST /ask` with tools (40 min)
• 6.1 Add the agent to `service.py`
• 6.2 Route `/ask` — one new route, that's it
• 6.3 Test the agent with a direct invocation
• 6.4 Expose `/ask` in the API Gateway
• 6.5 Agent tests (without the LLM, as always)
• ✅ Phase 6 Checkpoint
• 6.6 The same agent, with LangGraph (30 min)
• 6.7 Better retrieval: hybrid search (BM25 + semantic) (25 min)

• Phase 7 — Caching, structured logs and cost per invocation (25 min)
• 7.1 Structured logs and price per call
• 7.2 Cache-aside: don't summarize what didn't change
• 7.3 See the savings live
• ✅ Phase 7 checkpoint

• Phase 8 — Security and evaluation (30 min)
• 8.1 Red-teaming your own system: prompt injection
• 8.2 The guardrail: separating data from instructions
• 8.3 Golden dataset: does the agent answer well?
• 8.4 LLM-as-judge: when "contains the word" isn't enough (20 min)
• ✅ Phase 8 checkpoint

• Phase 9 — Auth and rate limiting (30 min)
• 9.1 Auth with JWT: who are you?
• 9.2 Rate limiting: how many times?
• ✅ Phase 9 checkpoint

• Phase 10 — Infrastructure as Code with Terraform (40 min)
• ✅ Phase 10 checkpoint

• Wrap-up — What transfers to real AWS (10 min)
• The bootcamp practices this project applies
• What this project DOES cover (and which week it comes from)
• What it does NOT cover (and exactly why)
• Where to go next

• Common errors and troubleshooting

No limit. It's a free guide: come in whenever you like, as often as you like.

No. Modules run from easier to harder, but you can jump to the one you need. Progress is saved per lesson.

Whatever is needed is listed under “What you need to bring”, above. If nothing is listed there, you can start from zero.

In the Club's WhatsApp group, and every two weeks there's a live with an instructor where questions get worked through.

Yes. It's issued automatically once you finish every lesson, with a verifiable code you can share on LinkedIn.