Claude AI Engineer / Anthropic Platform Engineer
Open a video or resource link, watch it, then check its box to track your progress and unlock what comes next.
- Build on Claude deliberately: context, tool contracts and coding-agent workflow
- Connect Claude to tools, data and multi-step work with production-grade agent patterns
- Operate Claude systems with eval-driven reliability, memory, cost discipline and secure autonomy
Benchmark: Anthropic Applied AI Engineer / Applied AI Architect + Staff Software Engineer, Claude Code / Claude Managed Agents signals (current Sep 2026). Market refresh: Claude API and context engineering; Claude Code; MCP/tool design; Agent SDK and Managed Agents; skills/subagents; evals; memory; sandboxing; context/cost efficiency; production reliability and enterprise safety.
Who’s Hiring Right Now
US openings posted in the last 7 days at Anthropic — this role’s own benchmark employers — refreshed daily, as of 2026-11-01. Pulled directly from each company’s public job board; not exhaustive, and postings close fast — worth confirming the details on the employer’s own site before applying.
Enrichment
Build on Claude deliberately: context, tool contracts and coding-agent workflow
Claude platform, context engineering + model behavior
Coverage: Claude API / Prompt & Context Engineering / Model Selection / Structured Outputs
Claude Code engineering workflow + repository context
Coverage: Claude Code / CLAUDE.md / Hooks / Commands / Subagents / Software Engineering
Connect Claude to tools, data and multi-step work with production-grade agent patterns
MCP, tool design + agent-computer interfaces
Coverage: Model Context Protocol / Tool Use / APIs / Permissions / Agent Skills
Agent architecture, managed execution + long-running workflows
Coverage: Claude Agent SDK / Managed Agents / Multi-Agent Orchestration / Sandboxing / Memory
Operate Claude systems with eval-driven reliability, memory, cost discipline and secure autonomy
Evals, regression + observable agent quality
Coverage: Agent Evals / Rubrics / Replay / Failure Analysis / Cost-Latency-Quality
Memory, context efficiency + proactive production operations
Coverage: Memory / Dreaming / Context Compaction / Prompt Caching / Routines / Reliability
Role-Readiness Proof / Capstone
Build a production Claude application for a consequential multi-step workflow. Start with a typed API service and explicit system/context contract; add structured outputs, retrieval or enterprise context, and at least two tools through MCP; implement a Claude Code-assisted development workflow with repository instructions, tests and hooks; add a managed or self-hosted agent path with sandbox/permission boundaries and optional subagent delegation; create a replayable eval suite covering task success, tool correctness, safety and failure recovery; instrument latency, tokens/cost and traces; then run a staged release, failure drill and rollback with a concise architecture/runbook package.
Proof Rubric
- Demonstrate model/context judgment
- reliable structured outputs
- high-quality tool contracts
- correct MCP integration
- disciplined Claude Code usage with human review
- secure permissions and sandboxing
- grounded retrieval where needed
- eval-driven development and regression gates
- diagnosable traces
- context/cost efficiency
- safe recovery from partial failures
- measurable task success
- and clear tradeoffs between prompts, tools, skills, subagents, memory and autonomous execution