Overview
In this case study I describe how I designed and led the implementation of an enterprise-grade Payment Switch handling mobile money, carrier billing, wallet transfers and merchant payments across multiple channels.
The Problem
The client was processing large volumes of USSD, mobile app, and API-initiated transactions but suffered from:
- High latency during peak hours
- No unified orchestration layer
- Tight coupling between channels & backend systems
- Poor observability and no end-to-end traceability
The Architecture I Designed
I introduced a cleanly decoupled, event-driven microservices architecture with:
- Java 17 + Spring Boot microservices
- Kafka-based event backbone
- Stateless services behind a centralized routing layer
- Redis + Postgres for caching and durable storage
- Fault-tolerant retry and circuit-breaker patterns
Key Improvements
- Reduced median latency from 950ms → 120ms
- Increased throughput to 20K+ TPS sustained
- Achieved 99.995% SLA
- Cutting deployment rollback time from 10 minutes to under 15 seconds
Impact
The platform now supports millions of customers daily and has become the core payment engine for 3 major business units.