[REMOTE] AI Engineering

JD — AI Engineering (Vietnam)

About the company

Our client builds the software hotels use to serve, know, and win back their guests. Founded

in June 2020 and headquartered in Yokohama, we also run a subsidiary in Vietnam—this role's

home base. The mission: evolve Japanese omotenashi through technology.

• IFHA (Inbound-First Hospitality Alliance): our alliance of 400+ hotels, ryokans and resorts across Japan, plus 15+ facilities in Vietnam.
• Guest-facing SaaS: GEM (guest engagement), MenyGo (mobile ordering) and Smartphone Concierge — 1.6M+ cumulative system users.
• AI visibility: AEO (answer-engine optimisation) audits, AI Profile — a machine-readable hotel data directory built for AI systems to cite — and the AI Learner Dashboard.
• IT consulting: DX strategy, system design and offshore development in Vietnam with JP/EN bridge engineering.

Why now. Hotels have lost their guests three times: to OTAs (booking), to Google (discovery)

and to COVID (the human touch went digital). The fourth shift is AI becoming the interface that

decides who gets found at all. We are rebuilding around it — tooling that lets one consultant serve many hotels, and an operations function run on agents.

The role

AI Engineering is the CIAO's (Chief AI Officer's) right hand and the builder-in-chief of our agent and automation stack. You will turn the CIAO's roadmap into shipped, production systems — largely by directing AI coding agents rather than by hiring a large engineering team.

Just as important as shipping: you own the standard those agents work to. The instruction layer — AGENTS.md / CLAUDE.md, sub-agent definitions, skills, prompt libraries, review gates and evals — is a product in this company, and it is yours to build, version and keep healthy as models and tools change.

This is a deeply hands-on technical role, but we are not hiring output — we are hiring leverage.

Your own code matters less than the difference you make to everyone else's: the engineer who

ships twice as much because your instruction files are good, the consultant who stops doing a

task by hand, the non-engineer who can safely run an agent. You will be judged on team

throughput and quality, not personal commitment count. That makes written and spoken communication a core technical skill here, not a soft extra: most of your leverage is delivered as

writing that other people (and agents) act on.

Reports to: Chief AI Officer (CIAO), Japan HQ

Works with: Product & AI team, Ops, Sales, and client-facing consultants

Type: Full-time, senior leadership

Location: Vietnam

What you'll do

1. Build with AI coding agents, daily. Design, prompt, review and ship production code — internal tools, data pipelines, integrations and client-facing products — using AI coding agents as your primary engineering workforce. We use Claude Code today and stay deliberately tool-agnostic: Codex, Cursor, Gemini CLI, Copilot and whatever ships next are all fair game if they do the job better.

2. Own and maintain the agent instruction layer. Author and continuously maintain AGENTS.md / CLAUDE.md and equivalent instruction files, repo conventions, sub-agent definitions, skills, slash commands, MCP tool wiring and scheduled jobs — so that any agent, on any tool, produces work to the same standard. Treat these as versioned, reviewed, tested assets, not scratch notes.

3. Set and enforce the engineering standard for AI-built code. Define how we write prompts, tests, evals, code review and guardrails; define what an agent may do unsupervised, what needs a human gate, and what is never automated. Keep a measurable bar (review pass rate, escaped defects, rework) rather than a vibe.

4. Run the agent workflow as a system. Design the handoff pipeline from requirement → scoped agent task → PR → review → deploy. Keep context hygiene, repo safety (pre-checks, reversibility, rollback), and destructive-operation policy explicit and followed.

5. Scale the consultant. Build the tooling that lets one our consultant serve many hotels — audits, reports, data structuring and delivery automation.

6. Support AI Profile and AEO products. Build and maintain the data pipelines, schema and feeds behind our hotel data products.

7. Run the CIAO's execution. Translate strategy into scoped projects, estimates and weekly delivery; report progress, cost and risk back to the CIAO.

8. Measure and control cost. Track model and token spend across tools, pick the right model and tool for each job, and keep unit economics healthy.

9. Multiply the team. Train engineers, consultants and non-engineers to work with AI tools safely and well; turn repeated work into reusable skills, templates and plugins; run short reviews and working sessions that raise the whole team's level instead of routing every hard task through you. Aim to make yourself removable from the critical path.

10. Communicate across borders. Write the specs, decision records, weekly updates and risk/blocker reports that keep Japan HQ, the Vietnam team and client-facing staff aligned — clearly enough that a non-engineer can act on them and a senior engineer can argue with them. Say plainly what is decided, what is open, and what needs the CIAO's call.

11. Keep us current. Evaluate new agents, models, MCP servers and workflow tools on a regular cadence; recommend adopt / trial / hold with evidence, migration cost and risk — not hype.

What you bring

• 3-5+ years in software engineering, owning a product's technical direction.
• Hands-on production experience with AI coding agents — you have shipped real systems this way, not just experimented. Any tool counts (Claude Code, Codex, Cursor, Gemini CLI, Copilot, custom agent loops). Be ready to walk us through a repo you built this way.
• Evidence that you raise the standard, not just the output. Show us the instruction layer you wrote — AGENTS.md / CLAUDE.md, sub-agents, skills, prompt or eval libraries, review checklists — and explain how it changed quality, and how you maintained it as models changed.
• Working knowledge of the modern agent toolset: project memory / instruction files, sub-agents, hooks, slash commands, skills, and MCP servers — at least deeply in one ecosystem, with the judgment to port the ideas across tools.
• Strong fundamentals in at least one of TypeScript/Node or Python, plus SQL, REST/JSON APIs, Git and cloud deployment.
• Experience building LLM applications: prompt design, tool use, evaluation, RAG or structured data extraction.
• Judgment about when AI output is wrong — you review agent code as carefully as a senior engineer reviews a junior's.
• Strong communication, in writing first. You can turn a messy discussion into a scoped spec, explain a technical trade-off to a salesperson and to an engineer in the same day, and give review feedback that people act on rather than resent. English writing that needs no cleanup before it goes to HQ or a client.
• A track record of lifting a team, not just a codebase. Mentoring, enablement, standards or tooling you introduced that measurably changed how others worked — and evidence you can point to for it.
• Security and data-handling discipline: secrets, access control, client data and licensing boundaries.
• A bias for shipping: small team, fast cycles, clear written updates.

Language (required)

Business-level English is required. You will work directly with the CIAO and leadership in Japan, write technical specs and status reports, and speak with international partners and clients — your written English should need no cleanup before it goes to HQ or a client.

Vietnamese is a plus for working with local vendors and talent. Cantonese or Chinese is a plus. Japanese is not required.

Nice to have

• Experience with agent SDKs or model APIs (tool use, prompt caching, managed/hosted agents) — Anthropic, OpenAI, Google or equivalent.
• Built or published MCP servers, agent skills, plugins or open-source agent tooling.
• Experience running evals or regression suites over LLM/agent output.
• Hospitality, travel, OTA or hotel-tech background.
• Schema.org, structured data, SEO or AEO (answer-engine optimisation) experience.
• Experience working with teams in Japan, Hong Kong or across Asia.
• A public GitHub, portfolio or write-ups showing AI-first engineering work.

Location: Vietnam (Ho Chi Minh City or Hanoi preferred), remote-first with occasional travel to Japan, potential to move to an office.

Hours: Regular overlap with Japan working hours (JST is 2 hours ahead of Vietnam).

Compensation: Competitive, based on experience.

Interview process: 3 rounds (online and offline tests will be required)