AI vibe coder

Job Title: AI Vibe Coder

Designation: Software Engineer (AI-Assisted Development)

Department: Engineering / Product

Employment Type: Full-time

Location: Hybrid — Baghdad / Remote

About the role

We're looking for an AI Vibe Coder who ships working software primarily by orchestrating AI coding agents rather than hand-writing every line. You'll live inside tools like Claude Code, Cursor, and agentic IDEs, turning loose product intent into running features fast. The ideal person has strong taste, knows when the AI is confidently wrong, and can tighten a "good enough" generation into production-grade code.

What you'll do

1. Translate vague product ideas and Jira tickets into shipped features using AI-assisted and agentic coding workflows.
2. Prompt, review, and iterate on AI-generated code — accepting what's right, rejecting what's slop, and refactoring the rest.
3. Build and maintain prompt patterns, project context files, and reusable scaffolds that make the whole team faster.
4. Own quality: write/validate tests, catch hallucinated APIs, and make sure generated code actually does what the spec says.
5. Pair with PMs and designers to go from concept to working prototype in hours, not weeks.

What we're looking for

1. 2+ years writing real software in any modern stack (we care more about judgment than a specific language).
2. Hands-on experience with at least one AI coding tool (Claude Code, Cursor, Copilot, Windsurf, or similar) on non-trivial work.
3. Ability to read and debug code you didn't write — including code the AI wrote.
4. Solid fundamentals: version control, testing, basic system design, and reading error messages without panicking.
5. Strong written communication; prompting is just writing clear requirements.

Nice to have

1. Experience with agentic workflows, MCP servers, or building internal AI tooling.
2. A portfolio of things you actually shipped (side projects count).
3. Comfort with one of: web frontend, backend APIs, or automation/scripting.

Red flags we screen for

1. Merging AI output without reading it.
2. Inability to explain how the code works once it's written.