Senior AI Interaction Evaluator: Codex & Claude Code

G2i Inc.; Miami (FL); On-site; USD 138,000 - 276,000; Full time

Job summary

G2i Inc. is seeking a Senior AI Interaction Evaluator to assess how coding agents behave in real-world scenarios. You’ll focus on whether responses are sensible, and whether explanations reflect solid engineering judgment rather than mere syntax.

This role emphasizes engineering taste over production coding, requiring senior engineers who can provide direct, opinionated feedback and define what great interaction looks like when using tools like Cursor, Codex, and Claude Code.

Qualifications

• Staff/Principal-level engineer (or equivalent experience).
• Strong background in TypeScript/JavaScript or Python.
• Hands-on experience using OpenAI Codex, Claude Code, Cursor.
• Deep familiarity with modern AI-assisted dev workflows.
• Able to evaluate code without fully executing or deeply reviewing every line.
• Comfortable giving direct, opinionated feedback.
• High bar for what “good engineering” looks like.

Responsibilities

• Evaluate AI-generated coding interactions end-to-end.
• Judge whether outputs are useful, correct at a high level, and show engineering judgment.
• Assess the quality of explanations and reasoning.
• Distinguish between levels of response quality (e.g., 2 vs 4).
• Provide clear, opinionated feedback on what worked, what didn’t, and what felt off.
• Help define what great looks like when interacting with Cursor.

Skills

TypeScript/JS

Python

Codex

Claude Code

Cursor

AI workflows

Engineering judgment

Tools

Cursor

Job description

G2i Inc. is seeking a Senior AI Interaction Evaluator to assess how coding agents behave in real-world scenarios. You’ll focus on whether responses are sensible, and whether explanations reflect solid engineering judgment rather than mere syntax.

This role emphasizes engineering taste over production coding, requiring senior engineers who can provide direct, opinionated feedback and define what great interaction looks like when using tools like Cursor, Codex, and Claude Code.