Untitled topic

In 2020, Kuda faced an existential crisis: their third-party core banking platform was failing at 100k users. System downtime every Monday. Weekend complaints. Failed transfers. They had two options: watch their bank collapse, or build their own core banking system from scratch.

Eight engineers chose to build. They delivered NERVE in six months—during COVID lockdowns—while keeping the bank operational.

Some things I found interesting from the interviews:

• Foundation over features: First 2-3 months produced zero visible endpoints. Just contracts, schemas, infrastructure, and observability. The CTO kept asking "can I see any APIs?" The team held firm—and after the foundation, development accelerated exactly as promised.

• Built for 10X scale: At 150k users, they built for 10 million. The same system now serves 10M+ customers without hitting designed limits.

• The night before launch: Production environment mysteriously failed. Their solution? Promoted the staging environment to production and shipped. It's still running today.

• Customers noticed immediately: Backend work is usually invisible. Not this time—customers started tweeting about faster response times on launch day.

• 90% cost reduction: Compared to enterprise solutions from vendors like Temenos or Finacle. No licensing fees, no vendor lock-in, full control over features and data.

• Load testing saved them: Designed for 1,000 concurrent transactions/second. Initial tests were "disastrous"—they weren't reusing Redis connections. Fixed connection pooling, hit targets, went live.

The full article covers architectural decisions (microservices, custom micro-ORM, synchronous transaction core with async edges), concurrency handling via row versioning, team dynamics during COVID, and what it took to convince leadership to invest in "invisible work."

Free to read until Dec 10, 2025: https://africanengineer.com/i/kuda-nerve

Would love feedback from anyone who's worked on banking cores, payment systems, or similar infrastructure migrations.