prasanna@builds:~$ project-detail

Order Automation & Fulfillment Orchestration

Automated end-to-end order orchestration across Amazon, Walmart, Shopify, and internal ERP — routing thousands of daily orders with near-zero human intervention.

Overview

Every order is a promise.
At Razor’s scale — with tens of thousands of daily transactions flowing through Amazon, Walmart, Shopify, and in-house systems — keeping those promises required orchestration, not hope.

Our fulfillment operations were powerful but fragmented.
Each system spoke its own dialect: e-commerce channels, ERP, and warehouse software all moved at different tempos.
My job was to make them dance in sync.

The Challenge

Manual intervention was the default.
Ops teams manually created ERP orders, cross-checked payment statuses, and triggered fulfillment center requests.
Each step invited delay, inconsistency, and error.
In a multi-channel environment, even a five-minute lag meant duplicate shipments, canceled orders, or missed cutoffs.

The goal was to design a deterministic pipeline that could receive, normalize, and route every order automatically — end-to-end — without breaking when any one system failed.

The Solution

I built a modular order orchestration framework on top of async APIs and queue-based jobs.
Each channel integration funneled into a normalization layer, mapping platform-specific payloads to a unified schema.
From there:

  1. Orders were deduplicated and validated against ERP state.

  2. Fulfillment centers were selected dynamically based on SLA and stock.

  3. ERP orders and ASN labels were generated automatically.

  4. Every webhook and downstream call carried a correlation ID for end-to-end traceability.

The pipeline used Redis-based task queues with idempotent operations, meaning retries were safe and side-effect-free.
Failures rolled into a dead-letter queue with replay support, so the system healed itself faster than humans could intervene.

Grafana dashboards and runbooks gave ops full visibility into the pipeline: where orders sat, what failed, and what auto-recovered.

The Impact

  • Reduced manual touchpoints by 80%

  • Increased fulfillment accuracy to 98%+ across thousands of orders per day

  • Improved SLA adherence and visibility across global fulfillment centers

  • Standardized orchestration logic across channels and ERPs

The orchestration layer became the invisible backbone of Razor’s order flow —
quietly doing the boring, essential work of keeping promises at scale.

The Learning

Automation isn’t about removing people.
It’s about removing friction — so people can focus on the problems automation can’t see.
What I built wasn’t just code — it was leverage.
And leverage scales quietly.

Tech Stack

Python · FastAPI · Redis Queues · PostgreSQL · Webhooks · ERP Integrations · Grafana · S3 Logging

Status

Confidential Internal Build
Architecture and orchestration logic proprietary to Razor Group.

Runtime for AI Tools

A lightweight runtime that standardizes how AI agents define, execute, and trace tools, reliable by design, portable across frameworks, and auditable anywhere. Brings predictability, reusability, and observability to AI tool execution, the missing runtime between frameworks and production.

Sarah Chen

Tech Stacks:

© Prasanna

v24.10.2025