B2B SaaS & Business Automation

ClassyCrave – B2B Operations & POS Platform

Role: Lead Software Architect   //   Timeline: 2026

ClassyCrave – B2B Operations & POS Platform architecture

Core Metrics

  • Eliminated long-lived connection overhead via smart dynamic polling (15s active / 60s idle)
  • Zero invalid payload submissions achieved via deeply nested Zod schema validation
  • Maintained 100% idempotency against Meta's automated retry mechanisms

Technology Stack

Languages

TypeScript, SQL

Orchestration

Upstash QStash, Next.js Server Actions

Data Layer

Neon PostgreSQL, Drizzle ORM

Frontend

Next.js, TanStack Query, React Hook Form, Tailwind CSS

The Problem & Constraint

A comprehensive Point-of-Sale (POS) and operations platform for the restaurant industry, featuring a custom conversational ordering pipeline, live kitchen synchronization, and strict transactional data integrity.

System Topology

Driven by an event-driven architecture using the official Meta WhatsApp Business API. Incoming webhooks are verified via SHA-256 HMAC signatures, trapped for duplicates via DB constraints, and pushed to a background job queue (Upstash QStash) for processing. The Kitchen Display System utilizes TanStack Query with aggressive polling and tab-visibility detection to manage live state without WebSockets.

Engineering Tradeoffs

  • 01

    Designed a highly constrained database schema utilizing PostgreSQL check constraints to prevent race conditions (e.g., negative sub-totals).

  • 02

    Implemented a granular, JSONB-driven Role-Based Access Control (RBAC) system with immutable activity auditing and instant session revocation via version tracking.