Agentic Software Engineer (Harness Engineer)

Quick Facts

• Role: Agentic Software Engineer (Harness Engineer)

• Focus: Building production software using AI coding agents and responsible multi-agent orchestration

Description

You’ll develop a production-grade, AI-first e-commerce platform by delegating significant development work to coding agents you control. The role centers on designing harnesses and workflows, adding verification/feedback loops, and ensuring code quality and system consistency across multi-agent outputs. You’ll also manage agent autonomy and operating cost to keep token usage efficient and safe.

Responsibilities

• Develop a production system with AI agents

• Design harnesses and agent workflows

• Split work across multiple agents; set autonomy level and manage shared context

• Build gates, checkpoints, feedback loops, and automated verification

• Detect and control context/specification/architecture drift and maintain code consistency

• Decide where testing/static analysis/secondary-agent checks are sufficient vs. where human-in-the-loop is required

• Control agent cost (token usage, iteration count, context size, model choice)

• Improve the agent working environment so agents can safely and economically take on more development

Requirements

• Very strong software engineering skills (not a prompt engineer)

• Practical experience with agentic engineering and tools such as Claude Code, Codex, Cursor, or similar

• Understanding of well-designed agentic harnesses vs. ad-hoc “vibe coding”

• Ability to set gates, implement verification loops, and detect hallucinations/drift

• Experience managing context and coordinating multiple agents on larger tasks

• Ability to manage agent economics: measure/limit token burn, choose models per task, control iteration cost, and recognize unproductive loops

• Ability to ensure agents verify solutions and do not break other parts of the system

• Must understand software engineering fundamentals: architecture, testing, Git, CI/CD, and working with large existing codebases

• Must be able to work comfortably in a PHP backend

Benefits

• Not required to hand-write most code; focus is on delegating work to agents

• During recruiting, an AI-agent workflow task/repository will be used to evaluate your normal agent-based working process rather than raw typing speed