bezgoFresh
The Vendor Control System
bezgoFresh's vendors ran on phone calls and scattered Google Sheets — prices, availability and regional names in different tabs, nothing real-time. We gave them one coordinated platform for orders, menus and daily sales, without the back-and-forth.
Behind every customer order sits a chain of hand-offs, and each one created confusion — vendors waiting for phone calls, menus out of sync, orders confirmed twice. We built a role-based platform where every zone shares one source of truth, and preparation status stays synchronized with operations.
The Challenge
The Bottleneck
The business coordinated dozens of vendors across multiple zones using scattered Google Sheets — orders, prices, availability and regional names lived in different tabs. Nothing was real-time. Vendors waited for phone calls to know what to prepare, and multi-vendor orders had no coordinated response flow. The process failed to scale with the region's daily demand.
The Solution
The Reality
Instead of calling vendors for every order, we gave them the platform itself. Menus can be updated digitally, orders arrive automatically, and preparation status stays synchronized with operations. Orders dispatch externally only once every vendor has answered — without losing the regional character of the local menus.
"We didn't just digitize a spreadsheet; we engineered a real-time order system that turns scattered vendor workflows into one coordinated platform."
Under the Hood
Technical Execution
A breakdown of the core systems engineered to deliver reliability and scale.
Real-Time Order Orchestration
- Engineered multi-vendor order ingestion where a single order is resolved and split across vendors automatically.
- Built a WebSocket ingestion layer for external order entry with Server-Sent Events pushing NEW_ORDER live to each vendor dashboard.
- Implemented coordinated response tracking — orders dispatch externally only when all vendors respond, with per-item accept/reject and preparing-weight capture.
- Hardened ingestion with pg_advisory_xact_lock and Redis dispatch to prevent duplicate processing across concurrent workers.
Four-Layer Menu Architecture
- Modeled a layered catalog: MasterItem → RegionalName → VendorItem → VendorMenu as a single source of truth.
- Built zone-based regional names in Malayalam with primary/alternative roles and a vendor-submission → hub-approval workflow.
- Added size-driven pricing for fish (SMALL / MEDIUM / BIG) with unit overrides and a menu history system for snapshot, reset, and restore.
Type-Safe Scale & Security
- Leveraged strict TypeScript across the entire codebase with Zod validation on every ingestion path.
- Enforced role-based access control (SUPER_USER / HUB_MANAGER / VENDOR) with forced password change, session limits, and login rate limiting.
- Isolated data by zone, offloaded item images to Azure Blob Storage, and added performance indexes across all hot queries.
Want to built something like this?
reach out to me... →The Outcome.
bezgoFresh went from Google Sheets to a live platform showing real-time vendor orders, coordinated multi-vendor responses, and unified regional menus — built and shipped in 12 weeks.
- Real-Time Order Sync
- Coordinated Multi-Vendor Flow
- Unified Regional Menus
What I learned
This project taught me that coordination is the hardest engineering problem.
Real-time isn't speed — it's waiting for every vendor before the world hears about the order.
A shared source of truth only holds when every zone has a reason to trust it.