Moolah Wireless: Carrier-Grade Mobile OS Backend & Subscription Engine

Architecting a distributed Node.js microservice cluster on AWS for real-time mobile device quota enforcement, recurring carrier billing, and automated SIM provisioning at scale.

99.98%

Production Uptime SLA

< 65ms

Avg Quota Lookup Latency

100k+

Connected Mobile Devices

Upwork

Verified Client Commendation

The Mission

Moolah Wireless delivers customized mobile operating systems on Android tablets and smartphones bundled with telecom data connectivity. As Lead Backend Architect, I engineered the API core powering device authentication, SIM activations across telecom partner networks, balance deductions, and billing subscriptions.

Node.js Express.js MySQL Cluster Redis Hot Cache AWS EC2 / S3 / RDS BullMQ Idempotent Queue

Hard Engineering Bottlenecks

Race Conditions in Live Quotas: When thousands of active devices stream video simultaneously, synchronous SQL transactions locked rows and caused database CPU spikes.
Unreliable Carrier Webhook Endpoints: Upstream carrier activation APIs frequently timed out or returned duplicate event notifications, risking double-billing or dropped SIM provisioning.
Hardware Security & Spoofing: The backend needed to verify immutable hardware UUIDs and cryptographic device tokens to prevent rogue SIM reuse.

Architecture & High-Throughput Topology

Device Boot / Custom Android OS Ping │ (Signed HTTPS Request + Hardware UUID) ▼ [AWS Route 53 / Application Load Balancers] │ [Node.js API Cluster (Horizontally Scaled)] ├── [Redis In-Memory Hot Cache] (Sub-5ms Device Policy & Quota State) └── [BullMQ Worker Queue] (Asynchronous Event Ingestion) │ ┌─────────┴─────────┐ ▼ ▼ [MySQL Primary-Replica] [Carrier Webhook Engine] (Billing / Audit Logs) (Idempotent AT&T/T-Mobile Connectors)

Idempotent Webhook Processing

To eliminate duplicate carrier activations and subscription charges, all incoming carrier webhooks pass through an atomic Redis key lock (SET resource_id token NX EX 30). If a duplicate event arrives within a 30-second window, the worker returns an instant 200 acknowledgment without re-executing database mutations.

Measurable Outcome & Verified Review

The system scaled smoothly to support over 100,000 active devices with 99.98% uptime SLA and sub-65ms average quota lookup latency.

"Kuldeep is an experienced full stack software engineer and his contributions to our work at Moolah have helped us grow and scale our technology. We will continue to build and work with Kuldeep on projects into the future. He is also a great communicator and team leader."
— Moolah Wireless Leadership (Verified Upwork Contract Review)

Scaling high-throughput APIs or complex backends?

I architect distributed backends designed for predictable performance under heavy production load.

Discuss Your Project View All Case Studies