Claude Certified Architect Jobs: Titles, Skills and Domain Mapping

Claude Certified Architect jobs almost never appear under that name; the skills the CCA-F tests show up in postings titled AI engineer, agentic or applied AI engineer, platform engineer, solutions architect and developer advocate. To find them, search by what the role does (Claude, agents, MCP, tool use, evals) rather than by the credential.

Below is a map from the five exam domains to concrete job tasks, the phrases job descriptions tend to use, and how to present the credential without overselling it. No job counts or salary figures appear here because Anthropic publishes none for the credential and any number would be a guess. Exam details are as of mid-2026; confirm on the official Anthropic / Pearson VUE page before you register.

Which job titles use CCA-F skills?

Title	What the role mostly does with Claude	Domains that matter most
AI engineer / applied AI engineer	Builds LLM features and agents end to end: prompts, tools, evals, deployment	D1, D4, D5
Agent engineer / agentic systems engineer	Designs multi-step and multi-agent workflows, orchestration, guardrails	D1, D2, D5
Platform / DevEx engineer	Standardises Claude Code across teams, writes MCP servers, wires CI	D3, D2
Solutions architect (AI)	Designs client systems, writes reference architectures, reviews trade-offs	D1, D5, D2
Developer advocate / DevRel	Explains patterns, builds demos and sample MCP servers, gathers feedback	D2, D3, D4
ML / data engineer with LLM scope	Batch extraction, structured output pipelines, quality monitoring	D4, D5

The pattern: engineering-heavy roles lean on Domains 1, 2 and 4; team-tooling roles lean on Domain 3; anything with “architect” in the title leans on Domains 1 and 5.

What do these job descriptions actually ask for?

Reading a sample of AI and agent-engineering postings, the recurring requirements look like this (observations, not statistics):

• Experience building applications on a frontier model API, frequently naming Claude or the Anthropic API alongside others.
• Tool calling or function calling, and often the Model Context Protocol by name for integrating internal systems.
• Some form of agent framework or SDK experience, plus opinions on when not to use agents.
• Evaluation and observability: how you know the system works, how you catch regressions.
• Structured output and schema design for downstream systems.
• Cost and latency awareness, including batching and context budgeting.
• For platform roles: Claude Code or similar coding-agent tooling adopted across a team, CI integration, permission and safety controls.

Notice how closely that list matches the exam blueprint. That is not a coincidence; the credential was designed around the tasks people are actually hired to do. It is also why the badge is a supporting signal rather than the main event: postings ask for the experience, and the exam certifies the judgment.

How do the five domains map to job tasks?

D1 Agentic Architecture & Orchestration (27%). On the job: deciding whether a workflow needs a model-driven loop or a fixed pipeline, splitting work across sub-agents that do not share context, choosing hooks over prompt instructions when a rule must never be violated, and knowing when to resume a session versus start fresh. Interviewers probe this with “walk me through how you would design an agent that…” questions.

D2 Tool Design & MCP Integration (18%). On the job: writing tool descriptions that do not overlap, keeping the tool set small enough that routing stays reliable, returning structured errors instead of “operation failed”, scoping MCP servers per project versus per user and keeping secrets out of config. This is the domain most likely to show up in a take-home.

D3 Claude Code Configuration & Workflows (20%). On the job: owning the CLAUDE.md hierarchy for a repo or org, path-scoped rules, skills versus commands, plan mode for risky changes, and running Claude non-interactively in CI with -p and structured output. Platform and tech-lead roles live here.

D4 Prompt Engineering & Structured Output (20%). On the job: calibrating classifiers with few-shot examples, using tool_use to guarantee JSON shape, handling nullable fields so the model does not fabricate, running the Batch API for large extractions, and designing multi-pass review so a second instance is not biased by the first.

D5 Context Management & Reliability (15%). On the job: keeping case facts from degrading over long sessions, designing escalation triggers that are not sentiment-based, distinguishing an access failure from a genuinely empty result, and monitoring by stratified sample rather than aggregate accuracy. Architects and anyone accountable for a production agent are asked about this constantly.

How should you present the credential in a job search?

• Name it precisely: Claude Certified Architect – Foundations (CCA-F), Anthropic, with the month passed and the Credly link. Note that it is valid for 12 months.
• Attach a proof of work. One repo, one internal write-up, or one demo per domain you claim. A small MCP server with clean error handling says more than the badge.
• Use the domain vocabulary in your bullets. “Replaced prompt-level rules with PreToolUse hooks for destructive operations” or “cut tool misrouting by rewriting overlapping descriptions” reads as experience, and it happens to be exam language too.
• Do not lead with it. Lead with the problem you solved; mention the credential as the thing that makes the pattern repeatable.

What the exam is really testing, and why interviewers care

Exam questions are scenario-based and ask which design is the least bad under given constraints. That is the same shape as a system-design interview: there is no perfect answer, only trade-offs against cost, latency, correctness and safety. Candidates who can articulate why the runner-up option fails tend to do well in both settings. Practising with explanations, rather than just scoring, is the transferable skill.

Next step

Run the free mock exam and look at your per-domain results; the weakest domain is also the one to fill in on your CV. Then read the Domain 1 study guide and the Domain 2 study guide, the two areas job descriptions mention most often, and see a week in the role for what the work looks like day to day.