March 2026
The Role Doesn’t Have a Name Yet. Here’s What It Does.
There is no agreed-upon title for the engineer who orchestrates AI agents instead of writing code. Here is what the role actually does — and the job description we wrote from scratch.
In my last post, I argued that SDD shifts the value of an engineer away from stack depth toward architectural thinking and specification quality. Several people asked the same follow-up: okay, so what do you actually call that person?
Nobody agrees. “Prompt Engineer” undersells it — this isn’t about writing clever prompts. “AI-Augmented Developer” sounds like a product feature. “Tech Lead” is closest, but misses the operational layer. At NOSOTA, we landed on AI Orchestration Engineer — someone who runs a virtual engineering team where the implementors happen to be AI agents.
The title will evolve. The responsibilities won’t.
Here’s what this role actually does: it interviews stakeholders and turns ambiguous requirements into precise, unambiguous specifications. It decomposes features into scoped tasks with explicit context boundaries. It orchestrates Claude Code agents across parallel workstreams, validates their output against architectural intent, and catches what’s technically correct but conceptually broken. It maintains ADRs. It owns quality.
What it doesn’t do: write boilerplate. Memorize framework APIs. Specialize in a single stack.
The profile looks nothing like a traditional senior engineer job description. So we wrote one from scratch. Here’s the short version:
AI Orchestration Engineer — NOSOTA · Remote · Full-time
You think in systems before you think in code. You write specs, not just tickets. You’ve shipped something real with Claude Code, Cursor, or Windsurf. You know Clean Architecture and DDD — not the syntax, the thinking. You can turn a stakeholder interview into a system boundary diagram.
If your first instinct is to open an IDE before the spec is done — this isn’t the role. If it’s to ask three more clarifying questions — keep reading.
