For senior engineers planning a cert

AI certification exam guide: CCAR-F and AWS AIP-C01, domain by domain

Anthropic CCAR-F

The Claude Certified Architect – Foundations (CCAR-F) is Anthropic's official credential. The pass mark is 720 out of 1000 on a scaled score (range 100–1000). The format is scenario-based multiple choice — one correct answer of four — and you are presented with 4 of 6 scenarios.

The scenario structure matters for how you prep: you sit 4 of 6 scenarios, each a situation you reason through rather than an isolated trivia recall. The domain weights are confirmed against the official exam guide.

The 60-question / 120-minute figures are confirmed in Anthropic's official v1.0 exam guide (effective July 2026), alongside the pass mark, format, and domain weights above.

AWS AIP-C01

The AWS Certified Generative AI Developer – Professional (AIP-C01) is an official AWS exam. The pass mark is 750 out of 1000. It has 65 scored plus 10 unscored questions, in a mix of multiple choice and multiple response formats.

The Cloudflare reality

As of 2026 there is no official Cloudflare AI or Workers certification. Cloudflare's formal certifications are security-focused. If you see a "Cloudflare AI cert" presented as an industry credential, it isn't one — there is no official exam to point to.

How to study for each

Map study time to the weights — the blueprint is the syllabus.

• CCAR-F: D1 Agentic Architecture & Orchestration (27%) and D3 Claude Code Configuration & Workflows (20%) are the heaviest, with D4 Prompt Engineering & Structured Output (20%) right behind. Spend the bulk of your prep there; D5 Context Management & Reliability (15%) is the lightest. Practice reasoning through scenarios, not memorizing facts — the exam is scenario-based.
• AIP-C01: D1 Foundation Model Integration, Data Management, and Compliance (31%) dominates — nearly a third of the exam — followed by D2 Implementation and Integration (26%). Together they are over half the marks. Treat D4 (12%) and D5 (11%) as the tail; cover them, but don't over-invest.
• Either exam: the weights tell you where a missed domain costs you most. A blind spot in a 31% or 27% domain is far more expensive than one in an 11% domain.

For the opinionated take — whether either cert is worth your time, and the ROI for your destination role — see which AI certifications are worth it for a senior engineer. This page is the blueprint; that one is the verdict.

As-built: how the module weighting maps to these domains

These blueprints aren't just quoted here — we built a curriculum against them. docs/CURRICULUM.md is the backward-design record: each Anthropic CCAR-F domain (D1 27% through D5 15%) is assigned to specific Track C modules, and each AWS AIP-C01 domain (D1 31% through D5 11%) to specific Track D modules, so module depth roughly tracks domain weight — Track D carries the most modules of any track partly because D1 (Foundation Model Integration, Data Management, and Compliance) alone is 31% of that exam.

The rule we hold ourselves to: a weight only gets used to justify curriculum depth after it's checked against the official exam guide, never carried over from a marketing summary. We keep a separate synthesized credential (Cloudflare CAPA, tracked in content/certs.json) explicitly labeled not an official exam, because we've caught ourselves — and others — treating "we teach it" as equivalent to "there's an exam for it." See AWS's certification pages and Claude's tool-use documentation for the primary sources behind these domains.

Frequently asked