04 — Claude Platform Architecture
Overview: Know Claude's API surface, model selection trade-offs, cost optimization levers, and deployment options. This is table stakes: you must be able to recommend the right model for the right use case without hesitation.
What This Module Tests
- Model selection judgment — Can you recommend Opus 4.7 vs Sonnet 4.6 vs Haiku 4.5 given constraints (cost, latency, quality)?
- Platform knowledge — Do you understand prompt caching, token economy, structured outputs, tool use?
- Deployment clarity — Bedrock (AWS) vs Vertex (Google) vs first-party API? You should know the customer implications.
Learning Path
Recommended order:
| Step | File | Purpose | Time |
|---|---|---|---|
| 1 | 01-Claude-API-Surface | API endpoints, message format, parameters | 15 min |
| 2 | 02-Model-Selection-Matrix | Opus 4.7 vs Sonnet 4.6 vs Haiku 4.5: cost, latency, quality | 15 min |
| 3 | 03-Prompt-Caching-Architecture | How prompt caching works, ROI, gotchas | 15 min |
| 4 | 04-Tool-Use-and-Structured-Outputs | Agent patterns, function calling, JSON output | 15 min |
| 5 | 05-Deployment-Surfaces | First-party, Bedrock (AWS), Vertex (Google): tradeoffs | 15 min |
Total notes time: ~75 minutes
Practice Exercises
| Drill | Focus | Time | Rubric |
|---|---|---|---|
| 01-30-Second-Model-Selection | Customer says: "We need an LLM for customer support." Which model? Why? | 2 min | Reasoning, tradeoffs, business awareness |
| 02-Whiteboard-Prompt-Cache-ROI | Draw out the ROI on prompt caching for a BFSI doc search use case | 10 min | Math, assumptions, customer benefit |
| 03-Bedrock-vs-Vertex-vs-First-Party | Customer is on AWS. Should they use Bedrock or first-party API? | 5 min | Governance, data residency, cost, latency |
Goal: All 3 at Hire level minimum. Strong Hire means you can hold the line on what Claude can't do.
Code Labs
Hands-on API practice:
- 01-First-Claude-Call — Your first API call
- 02-Tool-Use — Function calling and tool definitions
- 03-Structured-Output — JSON schema enforcement
- 04-Prompt-Caching — Cache implementation and ROI measurement
- 05-Custom-Eval-Harness — Build evaluation framework for your customer's use case
Running BFSI Customer
- 01-Claude-Platform-Decisions-for-BFSI — What model, features, and deployment surface for each BFSI use case?
Knowledge Check
Interview-QA-Bank covers: - Model selection & trade-offs (technical questions) - Cost optimization (HM questions) - Deployment strategy (executive questions)
👉 Skim the Q&A Bank first. Mark your 5 weakest questions.
How to Work This Module
- ✅ Read all 5 Notes in order (~75 minutes)
- ✅ Run the 3 Drills — Time yourself; answers should be immediate and confident
- ✅ Run the 5 CodeLabs — Get the Claude API in your hands; experiment
- ✅ Review the case study — Apply your knowledge to real BFSI scenarios
- ✅ Skim Interview-QA-Bank — Mark your 5 weakest questions
- ✅ Mark complete when all Drills reach Hire level
Strong-Hire bar: You can explain token economy math and prompt caching ROI in customer language.
What's Next
After this module: - You can recommend Claude confidently in any conversation - You understand cost and latency trade-offs - You've written code that calls Claude - You know deployment strategy for Indian BFSI (Bedrock India, Vertex India, data residency)
Next Module: 05-Solution-Architecture-Patterns
Quick Reference
Key Files: - Notes: Notes/ (5 files) - Drills: Drills/ (3 exercises) - CodeLabs: CodeLabs/ (5 implementations) - Case Study: Case-Study/ - Interview Q&A: Interview-QA-Bank