Architecture background
Phase 6 — System Architecture

Sovereign-Grade Payment Infrastructure

Six-layer microservices architecture built for Central Bank licensing, 10,000+ TPS capacity, and sub-400ms end-to-end latency across Africa.

PCI DSS Level 1 Central Bank Licensed 6 Countries AI-Powered

99.99%

Uptime SLA

+0.01%

< 400ms

Avg Latency

-12ms

10,000+

TPS Capacity

+2,000

3 Regions

Data Centres

+1 planned
Core Architecture

Six-Layer Microservices Stack

Each layer is independently scalable, deployable, and auditable. Click any layer to explore its constituent services and technologies.

Zero-downtime deployments via blue-green strategy
Event-driven communication via Apache Kafka
Circuit breaker patterns for fault tolerance
Horizontal auto-scaling per service
Transaction Pipeline

End-to-End Payment Flow

Watch a payment traverse the entire stack — from customer tap to merchant confirmation in under 400 milliseconds.

Real-Time Transaction Flow

End-to-end payment processing in < 400ms

0msTotal: ~400ms

Eight Layers. One Coherent Stack.

Each layer is independently deployable, horizontally scalable, and communicates via Apache Kafka event streams.

1

IDENTITY & ACCESS SERVICE

2

WALLET & LEDGER SERVICE

3

PAYMENT ROUTING ENGINE

4

ACCEPTANCE ENGINE

5

COMPLIANCE ENGINE

6

RISK & FRAUD ENGINE

7

REGULATOR DASHBOARD

8

INTEGRATION SERVICE

Designed for a Billion Transactions

Built to scale from launch to 10 million users without architectural rebuilds.

LAUNCH (Phase 1)
10,000–100,000 users
  • 3 pods per service, auto-scaling to 20
  • Single AWS region: af-south-1 (Cape Town)
  • RDS Multi-AZ with read replica
  • 3-node Kafka cluster, 12 partitions per topic
  • Redis 3-node cluster
Target:
500 TPS sustained, 2,000 TPS burst
GROWTH (Phase 2–3)
100,000–1,000,000 users
  • Multi-region: af-south-1 + eu-west-1
  • Database: country-partitioned, read replicas per country
  • PgBouncer connection pooling
  • Kafka: 6 brokers, 24 partitions
  • TimescaleDB for analytics workloads
Target:
5,000 TPS sustained, 20,000 TPS burst
SCALE (Phase 4+)
1,000,000–10,000,000 users
  • Aurora PostgreSQL Global Database
  • Active-active multi-region (Africa, EU, Asia)
  • Kafka: 12+ brokers, 48 partitions per topic
  • ElasticSearch for transaction search
  • Cold storage: S3 Parquet via AWS Glue
Target:
50,000 TPS sustained, 200,000 TPS burst

The Technology Stack

Every choice made for reliability, security, and scale.

BACKEND SERVICES

  • Node.js 20 LTS · TypeScript · Fastify 4
  • Prisma ORM · Zod validation · Jest testing
  • OpenAPI 3.0 specifications

DATA LAYER

  • PostgreSQL 16 · Redis 7 · TimescaleDB
  • AWS RDS Multi-AZ · PgBouncer
  • Flyway migrations

MESSAGE STREAMING

  • Apache Kafka · AWS MSK · MirrorMaker 2
  • 12 topics · 12–48 partitions · 3x replication
  • 7-day hot retention · S3 Glacier cold archive

INFRASTRUCTURE

  • AWS EKS (Kubernetes) · Docker · Terraform
  • Kong API Gateway · AWS CloudFront · Route 53
  • AWS MSK · AWS S3 · AWS KMS · AWS X-Ray

SECURITY

  • HashiCorp Vault · AWS KMS · AWS WAF
  • Argon2id password hashing · AES-256-GCM
  • RS256 JWT signing · TLS 1.3 minimum
  • Hardware Security Module (HSM) — Phase 2
Defence in Depth

Five Security Layers

Every transaction passes through five independent security layers. Each layer has its own scoring engine and can independently block suspicious activity.

Overall Security Score

Composite of all layers

96.8/ 100
Connected Ecosystem

External Integrations

Direct API connections to banks, mobile money operators, card networks, and identity providers across Africa.

15+

Banking

  • Stanbic Bank
  • Access Bank
  • Ecobank
  • Standard Bank
  • Zenith Bank
8

Mobile Money

  • MTN MoMo
  • Vodafone Cash
  • Airtel Money
  • M-Pesa
  • Orange Money
4

Card Networks

  • Visa
  • Mastercard
  • Verve
  • UnionPay
6

Identity

  • Smile Identity
  • Onfido
  • Ghana Card (NIA)
  • NIN (Nigeria)
Engineering

Technology Stack

Purpose-built with modern, battle-tested technologies chosen for performance, security, and regulatory compliance.

Backend

Go

Core services, ledger, wallet

Node.js

API gateway, orchestration

Python

ML models, data pipelines

Rust

Cryptographic operations

Data

PostgreSQL 16

Primary relational database

Redis Cluster

Caching & session store

Apache Kafka

Event streaming platform

TimescaleDB

Time-series analytics

Infrastructure

AWS (EKS)

Primary cloud provider

Kubernetes

Container orchestration

Terraform

Infrastructure as code

ArgoCD

GitOps deployment

Security

HashiCorp Vault

Secrets management

AWS HSM

Hardware key management

Falco

Runtime security monitoring

Snyk

Dependency vulnerability scanning

Design Philosophy

Architecture Principles

Event-Driven

Asynchronous event sourcing ensures every state change is captured, auditable, and replayable for regulatory compliance.

Zero Trust

Every service authenticates every request. No implicit trust between layers — even internal services use mTLS.

Horizontal Scaling

Each microservice scales independently based on load. Auto-scaling policies handle traffic spikes automatically.

CQRS + Event Sourcing

Command and query separation with immutable event logs. Perfect audit trail for Central Bank inspections.

Multi-Region

Active-active deployment across 3 AWS regions. Data residency compliance per jurisdiction with geo-routing.

Chaos Engineering

Regular failure injection tests ensure the system degrades gracefully. 99.99% uptime SLA backed by architecture.

Ready to Build on AuraTap?

Request detailed technical documentation or schedule a call with our engineering team to discuss integration.