Cross-platform normalization
A shared adapter layer translated SFCC OCAPI/SCAPI, Magento REST, HCL SOAP/BOD and Shopify GraphQL into consistent product and category operations.
01Selected team experience
Multi-Tenant AI Commerce Merchandising Platform
- Role
- Full-Stack / Platform Engineer
A multi-tenant merchandising platform across four commerce engines — giving retail teams one account-scoped workspace for product assortment, category sequencing, analytics, automation and AI-assisted merchandising.
- React
- TypeScript
- AWS Lambda
- Auth0
- DynamoDB
- SQS
- Step Functions
- OpenAI
- Terraform
- CloudFront
02Challenge
The problem
Merchandising work was fragmented across platform-specific tools, separate analytics vendors and manual category-ranking processes.
- Present one merchandising experience across multiple commerce platforms
- Support multiple accounts, stores, locales and catalog contexts
- Combine product, behavioral, social and inventory signals
- Automate category sorting and scheduled merchandising changes
- Move long-running and high-volume operations off the interactive request path
- Enforce authentication, account roles, feature flags and entitlements
- Deploy safely across development, QA, UAT and production
03Solution
How it was solved
A normalized commerce layer connected SFCC, Magento, HCL and Shopify. Account-scoped identity and entitlements protected tenants, while queues and orchestration moved long-running catalog operations away from interactive requests.
04Evolution
Delivery path
01 · Authenticate & select account
Auth0 login and entitled-account selection established the user’s account-scoped roles, stores and feature access.
02 · Choose commerce context
Merchandisers selected store, locale, catalog and category before loading normalized product data.
03 · Browse large catalogs
Virtualized grids and progress-aware pagination kept category exploration responsive at enterprise catalog scale.
04 · Sequence products
Drag-and-drop, bulk moves, clipboard actions, pinning, locking and explicit positioning controlled category presentation.
05 · Configure automation
Business rules, linked categories and recurring schedules coordinated automated ranking and synchronization jobs.
06 · Analyze performance
Traffic, conversion, revenue, geography, KPIs and executive summaries combined analytics from multiple providers.
07 · Apply AI intelligence
Similarity clustering, substitutions, bestsellers, trending products, frequently-bought-together suggestions and narrative KPI reports supported decisions.
08 · Prewarm delivery
Scheduled and manual crawler jobs populated caches before high-demand browsing and merchandising sessions.
05Architecture
Platform topology
API Gateway authorizers protected the edge; application hooks enforced account-scoped roles. Human users authenticated through Auth0 with retained OIDC options, while machine-to-machine credentials supported service calls. Terraform reproduced the platform across development, QA, UAT and production.
Merchandising SPA / widget Admin onboarding SPA
│ │
├──────────┬──────────────┤
▼ ▼ ▼
Auth0 identity API Gateway + JWT authorizers
│ │
▼ ▼
Shared auth library Lambda domain services
├─ config + entitlements
├─ rules + linked categories
├─ analytics + AI pipelines
├─ alerts + async requests
└─ cache-prewarming control
│
▼
Commerce adapters: SFCC / Magento / HCL / Shopify
│
┌─────────────────┬───────────────┴──────────────┐
▼ ▼ ▼
Commerce APIs Analytics providers AWS data plane
DynamoDB / S3 / SQS
Redis / Step Functions
EventBridge / Fargate06Product
Product surface
Catalog merchandising
Virtualized grids, progress pagination, drag-and-drop ordering, bulk moves, pinning, locking and explicit placement
Product presentation
Variants, colors, leading images, badges, product groups, attributes, facets and content slots
Catalog operations
Add, copy, replace, remove, publish/unpublish, refresh products and synchronize linked categories
Rules & automation
Sorting rules, recurring schedules, category jobs and Step Functions orchestration
Analytics
Traffic, conversion, revenue, geographic, KPI and executive-summary dashboards with CSV export
AI merchandising
Embeddings, hierarchical clustering, similar products, substitutions, bestsellers, trends and frequently bought together
Inventory alerts
Product watchlists and low-stock, out-of-stock and back-in-stock transitions
Platform operations
Feature flags, account entitlements, multilingual UI, cache warming and embedded documentation assistant
07Engineering
Technical highlights
Responsive enterprise catalogs
React Window virtualization, progressive pagination, Redis-backed caching and prewarming workflows kept large category operations usable.
Asynchronous command model
A 202-and-polling façade, SQS, Step Functions and EventBridge moved long-running updates and scheduled automation away from synchronous UI requests.
Account-scoped tenancy
Store configuration, Auth0 role claims, application hooks, feature flags and entitlements controlled both UI surfaces and API capabilities.
AI product intelligence
Embedding and hierarchical-clustering services generated similarity groups and substitutions, while asynchronous OpenAI analysis produced narrative KPI reports.
Embeddable delivery
The primary React product could run as a standalone SPA or custom-element widget using Shadow DOM isolation and extracted styles.
08Integrations
Connected systems
Commerce engines
Adapters supported Salesforce Commerce Cloud, Magento, HCL Commerce and Shopify through their respective REST, SOAP/BOD and GraphQL interfaces.
Analytics providers
A gateway normalized Google Analytics 4, Adobe/Omniture and retained legacy analytics sources behind one cached application contract.
Auth0 & OIDC
Auth0 SPAs, APIs, M2M clients, roles, claims, enterprise SSO and attack protection formed the primary identity layer, with Keycloak and generic OIDC compatibility retained in the UI.
OpenAI
Asynchronous KPI analysis produced narrative summaries from normalized performance data.
AWS data plane
DynamoDB, S3, SQS, Redis, Step Functions, EventBridge and ECS Fargate supported storage, queues, caching, orchestration and crawler workloads.
Delivery platform
Bitbucket Pipelines used AWS OIDC and Terraform to deploy private S3/CloudFront applications, Lambda services and environment-specific infrastructure.
09Scope
What the team delivered
Unified merchandising across four commerce platforms
Supported multiple accounts, stores and locales
Automated long-running catalog operations
Combined product, inventory and behavioral signals
Added AI clustering and decision support
Established controlled multi-environment delivery
10Services
Repos & services
Anonymized service family across the platform.
| Service | Responsibility | Stack |
|---|---|---|
| Merchandising Experience | Primary catalog, category, analytics and AI workflows as a SPA or embedded widget | React 17, TypeScript, Redux Saga, Material UI |
| Operator Onboarding | Internal account selection and Auth0-backed merchandiser registration | React, Auth0 |
| Identity Foundation | Tenants, applications, APIs, SSO, roles, claims, M2M tokens and shared auth helpers | Auth0, OIDC, Terraform |
| Commerce Integration Layer | Normalized product, category and badge operations across commerce engines | SFCC, Magento, HCL, Shopify adapters |
| Configuration & Rules | Account/store settings, entitlements, sorting, automation and linked categories | AWS Lambda, DynamoDB, Step Functions |
| Analytics & Intelligence | Provider normalization, caching, KPI narratives, social signals and AI clustering | Lambda, Redis, OpenAI, embeddings |
| Async Operations | Long-running request façade, jobs, queues, alerts and stock-transition processing | SQS, EventBridge, Step Functions |
| Cache Prewarming | Scheduled/manual crawling and operational monitoring for warm catalog caches | ECS Fargate, EventBridge |
11Outcomes
Results that matter
Unified merchandising workflows across four commerce engines
Supported multi-account, multi-store and multi-locale operation
Automated long-running catalog updates through asynchronous workflows
Combined product, inventory, behavioral and AI-derived signals
Established controlled multi-environment delivery through infrastructure-as-code and CI/CD
The represented product spanned 14 repositories across applications, domain services, identity, adapters and AI pipelines on a multi-environment AWS serverless platform.
Enterprise merchandising, multi-tenant SaaS architecture, commerce adapters, asynchronous workflows, AI product intelligence, identity and serverless cloud delivery.
12Boundaries
Scope & caveats
These implementation boundaries keep the public description technically honest.
- This was a team-built platform across 14 repositories; the role should not imply sole ownership of every subsystem.
- Some analytics providers and adapters reflected staged integration maturity rather than identical production depth everywhere.
- Compatibility libraries and identity adapters evolved with the platform’s multi-engine requirements.
- No conversion, revenue or merchandising-lift metrics are claimed for this case study.
13Visuals
Product visuals
14More work
Related experience
Virtual World Platform — Backend & Live Operations
A real-time backend and live-operations platform for a persistent virtual world — combining player progression, content releases, analytics and support operations across Go, NestJS, PostgreSQL, DynamoDB, Redis, NATS and Kafka.
View case study →
AI Outreach and CRM Automation
AI voice outreach connected to campaign operations and CRM follow-through — coordinating lead intake, structured outcomes, CRM updates, calendar booking and email follow-up across Vapi, Pipedrive, Twilio, FastAPI and AWS.
View case study →
Resilient Browser Video Platform
A resilient browser recording platform that lets users correct mistakes without starting over — coordinating screen, camera and audio capture, incremental upload, cloud processing, reversible editing, sharing, teams and billing.
View case study →
Related Peerprise services
15Next step