Work / 2026
Airwallex Billing
Designing a new billing platform from scratch for a late-stage fintech — where the hard work was getting the object model right for where enterprise billing is actually heading.
Context
Airwallex had grown to $1B+ ARR on billing infrastructure that was never designed to scale. Enterprise deals were complex — multi-product, multi-jurisdiction, combinations of real-time and post-billed payment flows — and the platform handled this through a mix of manual engineering workarounds and one-off configurations. Every bespoke deal added more surface area to maintain.
The business needed a proper billing platform. That became Airwallex Billing, launched in June 2026.
Problem
The immediate problem was operational: 80+ enterprise customers, each with bespoke billing arrangements, and no systematic way to manage them. But underneath that was a harder design question — what should a billing platform for a late-stage fintech actually look like?
Most billing platforms are built for SaaS subscription models. Airwallex’s reality was different: cross-border payment fees calculated in real time, usage-based billing on transaction volume, post-billed invoicing for enterprise contracts, and regulatory requirements varying by jurisdiction. Copying what Stripe Billing or Zuora had built would solve the wrong problem.
What I shaped
My main contribution was the object model — the set of core abstractions that the platform would be built on.
The key design question was: what does enterprise fintech billing actually need, not just today but as the business scales and as billing models evolve? Usage-based billing on event streams, tiered pricing across products, multi-currency fee calculation, audit-ready invoicing — all of these needed to be first-class concepts, not bolt-ons.
I worked closely with engineering on the model, stress-testing it against existing customer contracts and against the deal structures we expected to win in the next 18 months.
Key decisions and trade-offs
Design for configurability, not simplicity. Enterprise billing is inherently complex. The temptation is to simplify the model and handle exceptions manually. We did the opposite — built a model expressive enough to hold the complexity, so the exceptions could become configurations rather than code.
Map existing customers onto the new model before committing. With 80+ live enterprise customers, we couldn’t design in a vacuum. We systematically mapped their billing arrangements onto the new object model to surface where it broke, before we built anything.
Define migration paths as part of the design. A new platform is only useful if customers can get onto it without disruption. Migration paths were scoped alongside the object model, not after.
Don’t optimise for current deal structures. The model needed to support where billing was heading — UBB on event streams, metered usage, complex discount stacking — even if most current customers weren’t using those yet.
What launched
Airwallex Billing launched in June 2026. The platform supports fee calculation, invoicing, and financial reconciliation across global markets, with the object model as the underlying structure.
The operational impact was significant — 90% of manual billing workflows eliminated across enterprise customers, and the Commercial team can now configure complex deals without engineering involvement.
What I’d do next
The object model supports usage-based billing on event streams, but the ingestion and processing layer for those events is the next frontier. Getting UBB right at scale is the unsolved problem.