Summary
✨ AI‑Generated
Join an early platform engineering team focused on making software development faster, safer, and more automated. You will build developer workspaces, bootstrap environments, machine images, and reusable templates while improving the infrastructure that enables engineers and AI-assisted development systems to ship effectively.
Highlights
Remote role with direct collaboration with technical leadership, ownership of developer infrastructure, and an opportunity to shape an early-stage engineering platform.
Description
📍 Location: Remote (Decent overlap with HKT or EST working times preferred)
🌟 Job Type: Full-timeExperience Level: Mid to Senior (4+ years)
❖ About the Role
Most companies are adding AI tools to an existing engineering process.
Milton is built the other way around: coding agents write most of our code, and the engineering team's job is to make that fast and safe.
We call this harness engineering: making our systems legible to agents, and making safety enforceable rather than advisory, so agents can plan, build, review, and verify their own work while humans direct instead of type.
You'll be one of the first dedicated engineers on this platform, working directly with the CTO.
Your customers are our engineers and our agents, and your job is to keep raising the ceiling on how much both can ship.
❖ What you'll work on
The developer platform: cloud development workspaces (Coder on AWS), one-command laptop bootstrap, golden machine images (Packer), and the templates every engineer and agent works inside.The agent harness: Claude Code configuration treated as a product.
Skills, hooks, subagents, multi-agent design and code review pipelines, eval loops with machine-readable pass/fail signals, and the guardrails that make unattended agent work trustworthy.Infrastructure as code: a Terraform monorepo covering AWS, Cloudflare, Tailscale, GitHub org settings, and SaaS configuration.
Everything lives in code, every change is reviewed, applies are deliberate.Full-stack product and internal tools: MCP servers that connect agents to company systems (Feishu, Microsoft 365), internal bots and dashboards, onboarding sites, and features across our venture repos (TypeScript/Node, Python, Next.js).CI/CD and the paved road: GitHub Actions pipelines, plan gates, policy checks, and the shared conventions template that propagates all of the above across every repo in the org.
❖ What success looks like
A new engineer goes from sealed laptop to a first merged, agent-built PR in under a day.Agents complete larger units of work unattended because the loops, evals, and guardrails you built let them verify themselves.Velocity keeps climbing while incidents stay boring: the harness catches what reviewers used to.
❖ Requirements
4+ years building software across the stack (TypeScript/Node.js, Python, or similar, with some frontend).You already work through coding agents (Claude Code, Codex, Cursor) as your primary mode, not as autocomplete, and you have opinions about what makes agents effective.Hands-on infrastructure-as-code experience, Terraform strongly preferred, plus real cloud experience (we run on AWS).CI/CD fluency (GitHub Actions or similar) and git as a power user.Strong written communication.
In an agent-first org, documentation and conventions are load-bearing infrastructure; writing is part of the job.Security instincts: least privilege, secrets hygiene, blast-radius thinking.
❖ Nice to Have
Experience building developer platforms or internal tooling (Coder, devcontainers, monorepo tooling).Tailscale or other overlay networking; Cloudflare (DNS, Access, Tunnels).Experience building MCP servers, agent harnesses, or LLM-powered tools.Observability and incident response experience.Mandarin proficiency.
