eBay Next-Gen Billing Platform

At eBay I own architecture and delivery for key services in the company's next-generation billing platform, the system that meters, invoices, and reconciles what sellers across the marketplace are charged. The work sits at the intersection of financial correctness, high-traffic reliability, and operability: a billing error is not a cosmetic bug, it is money moving the wrong way, so the bar for correctness is absolute and the tolerance for downtime during change is close to zero.

The migration

The headline effort was moving the seller and balance data onto the new platform without disrupting a live, global seller ecosystem. Migrations of this kind fail in the seams, so most of the engineering was in the sequencing rather than the code: defining the cutover order, the validation gates each stage had to clear before the next could begin, and the rollback readiness that meant any stage could be reversed cleanly if a gate failed. The goal throughout was a cutover sellers never had to notice.

Public API rearchitecture

I rearchitected the public API layer that fronts billing at very high traffic. The focus was reliability under load: resilient request patterns, performance tuning on the hot paths, and rollout governance so changes reach production gradually and observably rather than all at once. The API is the contract the rest of the marketplace depends on, so backward compatibility and predictable behavior mattered as much as raw throughput.

Reconciliation and data pipelines

Underneath the platform I designed reconciliation and reporting pipelines built on deterministic validations, idempotent processing, and audit-friendly outputs, the properties a financial system needs to prove, after the fact, that every charge was correct. Alongside those I built and operated data pipelines across the warehouse and big-data platforms that feed billing analytics, reporting accuracy, and issue investigation.

AI enablement

More recently I have led AI-enablement work for the platform: modernizing the data contracts, observability, and governance an organization needs in place before AI-assisted workflows can be introduced safely and measurably. The premise is that AI belongs on top of a trustworthy data and reliability foundation, not instead of one.

How the work runs

Across all of it I drive the execution cadence, from effort estimation and prioritization through design reviews, and into production through incident triage and the follow-up hardening that keeps the same class of issue from recurring.