Fraud checks aren’t just for payments. Find out how to verify an Airwallex job or recruiter before you share your details.

View open roles

Fraud checks aren’t just for payments. Find out how to verify an Airwallex job or recruiter before you share your details.

Build the AI that runs global finance

Most AI job posts promise you’ll make people more productive. This one is different. You’ll build agents that move money, reconcile books, and run financial operations for 250,000+ businesses. The hard part is making them behave when permissions, approvals, auditability, and live customer workflows are on the line. Getting that right across 150+ countries and dozens of currencies is the challenge you’d take on.

View open roles
Businesses served worldwide
200 K+
Global payments processed annually
$ 250 B+
Bright and innovative people
1000 +
Offices worldwide
0
Nationalities represented
0 +

Hear it from the team

"We’ve learned that context beats instructions. Agents that understand why a change is needed produce better code than those following rigid specs."
Smiling image of AI leader
Andy
Engineering Director
"You have to be really comfortable with the unknown, because something we use today may be redundant tomorrow, so it's important to keep learning continuously."
Akankshita
AI Engineer
"The smartest model isn’t your moat. The infrastructure it runs on is."
Tim, Global VP of Data & AI
Timothy
VP Data and AI

See our open roles

Global finance is being rebuilt around AI, and the teams that get it right will shape how money moves for the next decade. If that’s what you want to build, check out our open roles.

ROLE
TEAM
LOCATION
View all open roles

This isn't a someday story, it's already shipping

We’re not asking you to imagine what the work could be. It’s in production, in front of customers today:

  • Kai — the AI assistant built into the Airwallex WebApp and expanding to where customers are. Customers use natural language to open accounts, enable payments, issue cards, and track transfers. Kai resolves 88% of conversations autonomously, with seamless human handoffs exactly when needed.
  • AgentOS — the finance infrastructure for the agent era. Skip the login and point any agent (Claude Code, Claude Cowork, Cursor) directly at your account. It executes live workflows—like reconciling last month’s transactions, paying a batch of invoices, or provisioning cards—entirely within your own tech stack and model.
  • T:0 — the AI-native finance platform built to scale from day zero. Specialized agents close the books, generate custom reports, and maintain a real-time financial model. Powered by a data graph fusing company context with modern accounting logic, T:0 is defining the standard for reliable financial agents where accuracy and trust are non-negotiable.

What you would actually work on

Getting a model to respond is easy. The engineering is in everything that happens once permissions, edge cases, and failure modes show up. Depending on your role, you could:

  • Build the agents behind Kai, T:0, and AgentOS — design how an agent closes month-end books, reviews expenses against policy in 100+ languages, or takes a customer from signup to their first payment.
  • Architect the agent runtime — define which tools a model can call, in what order, and what it is allowed to touch. Build orchestration for systems that are probabilistic by nature and unforgiving by requirement.
  • Design retrieval and memory that give an agent the context it needs — ground it in the right account, workflow, and policy so it can reason from facts rather than guess its way through a task.
  • Harden the guardrails — engineer the validation layers that intercept sensitive actions. Build rigorous eval frameworks to prove a model change actually made the system better, not just different.
  • Own observability for reasoning, not just requests — trace what the model saw, what it decided, which tool it fired, and where a workflow broke or handed off.
  • Ship frontier models into production — take a capability from “works in the notebook” to running live across reconciliation, operational support, and money movement at scale.

Who we're looking for

You’ll probably fit if you want:

  • production systems, not theoretical use cases
  • hard technical problems with serious constraints
  • architectural ownership, not prompt-tuning at the edges
  • work that ships and gets used by customers
  • room to influence the stack while it’s still taking shape


You’ll work alongside engineers who care about sound judgment, clean abstractions, and systems that hold up once they hit reality.

Why this team, specifically

You’re defining the playbook, not following one. Autonomous agents in regulated finance is genuinely unsolved territory. There’s no reference architecture to copy and no settled answer for how much an agent should be trusted to do on its own. You’ll make those calls and set the patterns other teams end up following.

The constraints make it better. You’re working where compliance, auditability, security boundaries, and operational risk are part of the problem from the first line of code. That raises the bar on how carefully you design, and it’s why the systems you build are ones people can trust with their money.

Live scale and room to build. Usually you compromise: either the product is in market but the defining decisions are locked, or it’s a greenfield with nothing shipping yet. Here you get both, systems in production and open ground to decide where they go next. You’re joining early, too: Airwallex is still climbing the value curve, so serious scope and meaningful pre-IPO upside come with the work.

How we hire

Our process is rigorous, transparent, and two-sided — you’re assessing us as much as we’re assessing you.

We care less about benchmark results and more about how you’ve used AI to solve real problems for real users. Does this sound like your next step?

Apply now