Coding Agent Engineer

Coding Agent Engineer About the Role

We are building an AI application community where anyone can create, and everyone can benefit. At the heart of this vision is our Coding Agent-the core infrastructure that enables creators with no engineering background to turn ideas into fully functional applications within minutes.

The mission of this role is simple:

Enable the Agent to generate code that works the first time, deploys successfully, and delights users.

You will own the entire agentic coding loop, from prompt tool call code editing sandbox build preview deployment, while driving its reliability, observability, and continuous improvement.

This is not about maintaining an existing system-you will help define its architecture, tool protocols, and quality standards from the ground up.

What You'll Build Tool System Reliability

• Design model-centric tooling for code reading, editing, search, build, and diagnostics.
• Continuously improve tool invocation success rates and first-pass code editing accuracy.
• Analyze Agent failure patterns in tool usage and systematically improve tool interfaces and behavior.

Agentic Loop Orchestration

• Build domain-specific sub-agents across multiple frameworks, including Next.js, Remix, Vite, and static HTML.
• Design orchestration strategies that ensure consistency across sub-agent outputs.
• Optimize build feedback loops, convergence speed, and idempotent execution.

Context & Memory Engineering

• Design long-term project memory schemas and context injection strategies.
• Manage project plans, recent changes, and engineering conventions while maximizing the effectiveness of limited context windows.
• Ensure the Agent always maintains the correct engineering context.

End-to-End Sandbox Infrastructure

Own the complete pipeline for:

• Code build
• Live preview
• Vercel deployment

Drive first-build success rates and deployment reliability across multiple framework templates to production-ready standards.

Automated Diagnosis & Self-Healing

Build an autonomous feedback loop capable of:

• Reading build/runtime errors
• Generating fixes automatically
• Validating application usability

Our goal is for the Agent to resolve the vast majority of engineering issues without human intervention.

Guardrails & Output Validation

Design production-grade guardrails including:

• Tool behavior constraints
• Source code write protection
• Deployment gates
• Security validation
• Reproducibility guarantees

Capability Expansion

Extend the Agent's capabilities by designing and implementing:

• MCP Servers
• Agent Skills
• CLI tools

What We're Looking For-Required Qualifications Deep Understanding of Agentic Coding

• Strong understanding of the complete Agentic Coding Loop:
  • Tool selection
  • Code editing correctness
  • Build feedback
  • Self-fix convergence
• Ability to analyze Agent failures from first principles and systematically improve reliability.

Prompt Engineering

• Extensive experience designing production-grade prompts for coding tasks.
• Experience with:
  • Multi-stage prompting
  • Self-repair feedback loops
  • Prompt optimization for build and deployment workflows.

Tool System Design

• Experience designing AI-facing tool interfaces, including:
  • Schema design
  • Tool descriptions
  • Return formats
• Hands-on experience building code editing, search, and build tools.

Multi-Agent Systems

• Experience designing and orchestrating domain-specific sub-agents.
• Ability to coordinate multiple agents while maintaining consistency across generated outputs.

Context & Memory Engineering

• Strong understanding of context window limitations.
• Experience designing memory architectures and context injection strategies for long-running AI workflows.

Front-End Build Systems

Strong experience with modern frontend ecosystems including:

• Next.js
• Remix
• Vite
• Tailwind CSS

Ability to diagnose and resolve common build and runtime issues.

Guardrails

Experience designing application-level AI guardrails, including:

• Tool restrictions
• Source code protection
• Output validation gates

Sandbox Infrastructure

Hands-on experience with:

• E2B
• Containerized execution environments
• Code execution sandboxes

MCP / Skills / CLI

Strong understanding of:

• MCP protocols
• Agent Skills
• CLI tool ecosystems

Programming Languages

Required:

• TypeScript
• Node.js

Preferred:

• Python
• Bun

Software Engineering Experience

• 3+ years of full-stack engineering experience.
• Experience shipping production applications built with Next.js, Remix, or Vite.

Product Quality Mindset

Strong attention to:

• Reliability
• Reproducibility
• Regression prevention
• Whether AI-generated applications truly work for end users

Preferred Qualifications

Candidates with multiple of the following will stand out:

• Engineering experience on AI coding products such as:
  • Claude Code
  • OpenAI Codex
  • Cursor
  • Devin
  • v0
  • Bolt
• Experience with programmatic code understanding technologies, including:
  • AST manipulation
  • Codemods
  • LSP Diagnostics
  • Symbol indexing
  • Reference analysis
• Experience building production governance systems for multi-framework templates, including:
  • Automated canary releases
  • Error budgets
  • Rollback mechanisms
  • Cross-template regression testing
• Experience building production-grade or open-source:
  • MCP Servers
  • Agent Skills
  • CLI tools
• Ownership of open-source projects or commercial products with outstanding engineering quality or user experience.