From Parse.js to CQRS: Rebuilding a YC Startup's Backend

The Challenge

BUMP (YC W18) had a problem that would have killed most startups: their entire backend was a single-file Parse.js application.

As a StockX-like sneaker and streetwear marketplace handling high-value transactions, they needed authentication flows, order verification, payment processing, and inventory management—all running through one file.

The CTO knew this wouldn’t scale. He recruited me as the first engineer on a new full-stack team to fix it—while the mobile apps continued shipping features on top of the existing system.

No pause button. No rewrite-and-replace. The migration had to happen live.

What I Did

Backend Migration

The first priority was getting off Parse.js without breaking the mobile apps:

• Migrated the entire backend from a single-file Parse.js app to NestJS + GraphQL
• Designed typed APIs with full validation to catch errors before they hit production
• Coordinated closely with the mobile teams to ensure seamless FE integration during the transition
• Set up proper testing infrastructure to prevent regressions

Event-Driven Order System

BUMP’s marketplace had a complex verification flow—like StockX, every high-value item needed authentication before the transaction could complete:

• Architected a CQRS-based event-driven order system to handle the multi-step verification process
• Implemented proper saga patterns to ensure consistency across payment, verification, and shipping states
• Built the system to handle edge cases: failed verifications, payment disputes, shipping delays
• Ensured eventual consistency across distributed services without losing orders

Team Leadership

Beyond the technical work, I helped build the engineering function:

• Managed SCRUM ceremonies and sprint planning for the team
• Led the end-to-end engineering hiring pipeline—from job specs to interviews to offers
• Established code review practices and deployment processes

The Tech Stack

• NestJS (Node.js)
• GraphQL (Apollo)
• CQRS + Event Sourcing patterns
• Saga-based transaction management
• PostgreSQL
• Redis (caching + queues)

The Results

• Single-file → Microservices — Complete backend architecture overhaul
• Millions — In marketplace transactions processed
• Zero — Data loss during live migration
• Full team — Engineering function built from scratch

The new architecture handled the transaction volume that would have crashed the Parse.js system. Orders flowed through verification, payment, and shipping with proper state management. The mobile apps never skipped a beat.

More importantly, BUMP now had an engineering foundation that could scale with the business—and an engineering team to maintain it.