Building ACE: Applying Financial‑Style Architecture to a Unified Cross‑Border E‑Commerce Campaign Platform

Rapid expansion of AliExpress’s e‑commerce business created fragmented campaign systems for each line (homepage, promotion halls, marketing channels, etc.), leading to duplicated development, high maintenance costs, and systemic stability risks.

ACE Overview

ACE (AliExpress Campaign Engine) is a unified campaign platform for eight major business teams. It provides standardized infrastructure that can be reused like building blocks while allowing each team independent development and release processes.

FY25 – Building the Foundations

In FY25, ACE completed class‑loader isolation, resource isolation, unified monitoring, and disaster recovery, effectively "repairing roads and building bridges" for the platform.

FY26 – From Foundations to a Full‑Scale System

FY26 shifts focus to constructing a complete "digital asset management system" inspired by financial architecture:

Gateway – Payment‑clearing hub : consolidates traffic entry, solves chain chaos and security issues.

Workbench – Business hall & risk‑control center : centralizes metadata configuration and intelligent diagnostics.

Standardization – Financial regulation : defines unified models, links, and campaign rules.

Intelligent – Quantitative investment engine : replaces manual decisions with AI‑driven end‑to‑end automation.

Chapter 1 – ACE Gateway

The gateway acts as the "clearing system" for all campaign traffic. Legacy issues included outdated architecture, multiple parallel gateways, and weak traffic‑protection (QPM > 9 million). The redesign introduced:

Brand reconstruction and logic decoupling via ace-gateway-s, isolating ACE’s brand mind.

Removal of heavy rich‑client components, bringing core logic back to the native application.

Layered design separating traffic entry, business capability, core routing, and infrastructure.

Enhanced stability with four‑layer traffic protection.

Results: request hops reduced from 3 to 1, operational manpower freed, and issue‑location time cut by over 50%.

Chapter 2 – ACE Workbench

The workbench serves as both a "business hall" for metadata configuration and a "risk‑control center" for intelligent diagnostics.

Metadata Management : visual CRUD for fields, supporting supplier and dataFetcher types, moving from manual Diamond JSON edits to UI operations.

Model Field Management : defines field availability per scenario (activity, gameplay, template version) and exposure controls.

Model Binding Management : links entity types, process IDs, and fields to form complete campaign models.

Intelligent Binding : AI‑assisted recommendations and rule identification reduce manual effort.

Impact: configuration release efficiency improved >70%, error rate dropped dramatically, and rollback cost reduced >70%.

Chapter 3 – Standardization Capability

Standardization establishes a unified "regulatory" framework:

Model Standardization : defines asset categories (entity types), underlying assets (DO), and product definitions (DTO).

Link Standardization : consolidates disparate front‑end views (portal, layer, coupon, static) into a unified execution flow.

Campaign Standardization : refines the smallest reusable unit from a whole page to a "point of placement" (module + data + schedule), enabling fine‑grained reuse across pages.

Chapter 4 – Intelligent Selection & Placement (AI‑Native)

Transitioning from a manual "assistant" (Version 1.0) to a full quant‑trading engine (Version 2.0) involves:

AI‑driven strategy parsing from natural language, Excel, or design drafts.

Automated page planning that outputs page metadata, structure, and floor‑level configurations.

Knowledge‑base layers (frontend, campaign, business) that constrain AI generation.

Automatic code generation (X2C) that turns AI‑planned pages into deployable code.

Point‑placement engine that auto‑creates targeting schedules, binds recall strategies, and feeds back performance data for continuous AI improvement.

The AI‑native vision aims to shift from "human‑driven, AI‑assisted" to "AI‑driven, human‑supervised", enabling real‑time perception, proactive risk mitigation, and end‑to‑end automation.

Future Outlook

FY27 will focus on agent‑based transformations, knowledge‑graph enrichment, and expanding AI autonomy while redefining individual technical competitiveness toward problem definition, system‑wide understanding, and human‑AI collaboration.