Module 1: Exam Orientation and Strategy

Lesson 1: Exam Structure and Scoring Model

The Claude Certified Architect – Foundations (CCA-F) is not a test of what you can remember; it is a test of how you think. In the world of AI architecture, "knowledge" is ephemeral—models change every six months. What remains constant is the Architectural Framework you use to evaluate constraints, trade-offs, and outcomes.

In this first lesson, we will deconstruct the "Anatomy of the Exam." To pass, you must understand not just the technical content, but the Scoring Engine that determines whether your design decisions meet the professional standard of a Certified Architect.

---

1. The Professional Context of CCA-F

The CCA-F is designed for individuals who bridge the gap between "Prompting" and "Production." While a developer might ask "How do I make Claude write this list?", an Architect asks, "How do I ensure this list is generated with sub-second latency, validated against a schema, and cached to reduce token cost by 80%?"

The exam structure reflects this shift. You are expected to operate at the System Level.

The Domain Weighting

The exam is divided into five core domains. Understanding the weight of these domains is your first strategic step. If you excel at prompting but fail at reliability, you will not pass.

Domain	Weight	Focus Area
Domain 1: Agentic Orchestration	25%	Multi-agent patterns, decomposition, and state management.
Domain 2: Tool & MCP Design	20%	JSON Schemas, protocol contracts, and side-effect safety.
Domain 3: Logic & Output Control	20%	Structured outputs, validation loops, and complex prompting.
Domain 4: Performance & Context	20%	Token economics, caching, SNR optimization, and RAG.
Domain 5: Reliability & SRE	15%	Fallbacks, semantic validation, and error recovery.

---

2. The Anatomy of a Question

Unlike basic certifications that use multiple-choice questions (MCQs) for factual recall, the CCA-F uses Scenario-Based Inquiries (SBIs).

The Three Parts of an SBI:

1. The Context: A real-world business or engineering pain point. (e.g., "A retail company is experiencing high latency in their customer support agent...")
2. The Constraints: Specific technical or budgetary limits. (e.g., "...they have a budget of $0.05 per interaction and require 99% accuracy.")
3. The Call to Action: A request for a specific architectural decision. (e.g., "Which strategy most efficiently reduces cost without sacrificing accuracy?")

Why this is challenging:

There is rarely a "Wrong" answer in the sense of being technically impossible. All choices might be valid Python code. The "Correct" answer is the one that fits the Constraint Pareto Front—the best balance of the specific limits provided in the prompt.

---

3. The Scoring Model: Logic over Luck

The CCA-F uses a Partial Credit and Logical Consistency scoring model for its more complex items.

Multiple-Response Selection

Some questions require picking "The Two Best" options. These are designed to test your ability to see the system as a whole. For example, reducing latency might require both a faster model (Haiku) and a better caching strategy. Selecting only one will result in partial points.

The "SCA Justification" Rubric

In later levels (Associate/Professional), you must write justifications. In Foundations, the justifications are built into the distractor logic. You are scored on your ability to Identify the Primary Lever.

• If a problem is caused by "Prompt Injection," the fix isn't "Better search logic"; it's "Input Delimitation."
• If the problem is "Infinite Loops," the fix is "Deterministic Exit Conditions."

---

4. Technical Baseline: What You Must Know

While this is a "Foundations" exam, the technical baseline is relatively high. You are expected to be familiar with the following frameworks, even if you aren't an expert coder in all of them:

• Python (FastAPI): As the primary glue for AI agents.
• JSON Schema: The "Source of Truth" for tools.
• MCP (Model Context Protocol): The standard interface for data.
• LangGraph: For stateful, graph-based agent orchestration.
• AWS Bedrock / Docker: For deployment and sandboxing.

Architectural Diagramming

You will be asked to interpret and occasionally select the correct Mermaid diagram for a given workflow.

graph TD
    A[Human User] -->|Query| B{Supervisor Agent}
    B -->|Task 1| C[Search Worker]
    B -->|Task 2| D[Coder Worker]
    C -->|Results| B
    D -->|Artifact| B
    B -->|Unified Response| A

In the diagram above, you should immediately recognize the Supervisor-Worker pattern (Module 3). If the scenario asked for "Distributed Autonomy," this diagram would be a "Distractor."

---

5. Summary and Next Steps

To succeed in the CCA-F, you must adopt the Architect's Mindset. Every decision has a price (latency, cost, complexity).

In the next lesson, we will dive into the Scenario-Based Reasoning Approach, learning the step-by-step algorithm for deconstructing a complex problem and arriving at the SCA-preferred answer.

---

Interactive Quiz

1. Which domain holds the highest weight in the CCA-F exam?
2. What are the three components of a "Scenario-Based Inquiry"?
3. Why is "Model Tiering" (Module 10) often the correct answer for budget-constrained projects?
4. What is the difference between "Deterministic" and "Probabilistic" scoring in AI certification?

---

Reference Video: