What we're trying to build

We're Finan — the team behind SoBanHang, a business app used by a lot of small shops across Vietnam. We're now building FinanOne, an AI-native financial OS for small and medium businesses.

The idea is simple: most Vietnamese SMEs can't afford a real CFO. We want to give them something close to one — an AI that quietly handles the operational accounting work, flags cash flow problems before they hurt, and helps owners get to the right banking product when they actually need it.

We're early. The product is real, paying customers are coming in, but we have a long way to go. The AI layer is the heart of what makes this different from a normal accounting tool, and right now it doesn't have someone senior owning it end-to-end. That's why we're hiring you.

What you'd be doing

You'd lead our AI work — architecture, model choices, evaluation, cost economics, and eventually a small team. To start, it's mostly you and a few engineers. By end of year one, we'd hope to grow that to about five people, if the work supports it.

The first six months are concrete:

• Ship one production AI agent that does something measurably useful for paying customers
• Build the evaluation framework so we stop deploying by gut feeling
• Get our Vietnamese intent and entity extraction working well enough to trust in production
• Set up the cost dashboard so we know what every AI call costs and why
• Hire two or three engineers to work alongside you

We'll write the detailed milestones together in your first two weeks. We have opinions but we want yours to override ours where you see better.

The hard problems are real

We don't have buzzwords to offer — just real engineering work:

• Vietnamese in a finance context. English-trained models don't know our accounting terminology, our tax structures, or how an SME owner actually talks. Someone has to figure out where fine-tuning helps, where prompting is enough, and where we just need better data.
• Cost economics that actually work. SME pricing is thin. If every AI call costs as much as a premium model, we don't have a business. Routing, caching, and model cascade have to be measured and defended, not hand-waved.
• Hallucination tolerance near zero for money. When the AI tells someone their cash position, it has to be right. Designing the human-in-the-loop policy and the audit trail isn't optional — it's the work.
• Multi-agent systems that occasionally disagree. One agent says approve, another says hold. You design how that gets resolved without dropping the customer into a black box.
  

Who we think fits

Honest version: we're a small team with limited runway, so we need someone senior enough to make decisions without a lot of supervision.

Most of these matter:

• Around 10 years of software engineering experience, with the last few focused on shipping AI to real users
• You've put LLM-powered products in front of actual customers — not just demos or research projects — and you've felt the pain of incidents, on-call, and weird production behaviour
• You've hired and managed a small AI/ML team before, even just three or four people
• You've built agentic systems with tool use, conversation state, retry — not just single-turn chat
• You have a real point of view on evaluation: how to know your AI is getting better instead of just feeling like it
• You've thought hard about cost at scale: model routing, caching, batching
• Python plus one production language (Go, TypeScript, or Java) — you can read and write both comfortably
• You can write a technical spec clear enough that other engineers can build from it without daily check-ins

Useful but not required:

• Fintech, healthcare, or other regulated-industry background
• Returning Vietnamese from a strong AI team abroad (Google, Meta, OpenAI, Anthropic, Scale, Cohere, or similar)
• Production experience fine-tuning models for a non-English language
• Time as a tech lead or founder at an earlier-stage startup — you know what early-stage chaos feels like
• Native or near-native Vietnamese, especially if you've worked around finance or accounting
  

What we can offer

We're a startup, so the honest answer is: less cash than a big company, more ownership and impact than you'll find anywhere else at our stage.

• Competitive base for the Vietnam / Singapore senior AI market. We don't ask about your previous salary. We quote based on level and role. Specific range gets discussed in round one.
• Meaningful equity with the standard four-year vest and one-year cliff. We'll explain the cap table honestly in round three.
• Performance bonus tied to year-one outcomes we agree on together.
• Relocation support if you're a returning Vietnamese candidate moving back.
• Visa support if you're coming from outside Vietnam.
• Personal LLM API budget to experiment, plus access to our internal AI tooling.
• Time with our investors and advisors — including Kevin Ryan (founder of MongoDB and DoubleClick) and Rob Frohwein (founder of Kabbage). They're genuinely involved, not just names on a deck.

What we can't offer: a finished product, a fully built team, or a roadmap that's already certain. If you want those things, we're probably too early.

How to apply

Send your CV to hainam@finan.one with the subject [Head of AI] Your Name.

Please include:

• CV (PDF, one to two pages is plenty)
• Links to one or two production systems you've shipped. Three lines on each: what it does, your specific role, and one thing that broke and how you fixed it
• A link to a technical spec, RFC, or architecture doc you've written
• Optional: a five-minute Loom introducing yourself and a project you're proud of

If your profile looks like a match, we'll get back to you within five working days. The whole process usually takes three to four weeks. If we make you an offer, it'll come within 48 hours of the final round.

We read every application. Even if we say no, we try to say it honestly and quickly.