Design for change
Frequent business variation belongs behind clear configuration boundaries and reusable components where that makes sense.
Lead Backend Engineer · Distributed Systems · Financial Platforms
Backend platforms for systems that need to scale, change and stay reliable.
12 years across digital payments, banking, lending, telecom and enterprise software. Current work centres on configurable financial platforms, partner integrations, performance engineering and production reliability.
12+years of engineering
110+concurrent campaigns
~8Meligibility evaluations / day
35banking partners
5×throughput improvement
8backend engineers led
Engineering approach
Look for the constraint behind the symptom.
Slow APIs are not always database problems. Threads, caches, logging, downstream services and JVM behaviour can all shape latency. The approach is to establish the evidence first, then change the system.
Measure before scaling
Traffic, JVM, threads, database, cache, logging and downstream latency provide a better basis for decisions than assumptions.
Keep operations simple
Good production design should be observable, recoverable and practical for the team operating it at 2 AM.
Selected work
A few production problems worth a closer look.
01 · Performance
40 → 200 RPS per instance.
JVM analysis, multi-level caching, database optimisation and asynchronous processing improved throughput while reducing the application footprint.
14 → 8 instancesRead →
02 · Production
When logging became the bottleneck.
Thread dumps and Grafana exposed request threads waiting in the logging path during peak traffic.
Threads · I/O · LogbackRead →
03 · Automation
Turning recurring campaign work into configuration.
A small scheduling model removed manual activation and deactivation from a recurring campaign workflow.
Daily · weekly · monthlyRead →
04 · Preventive
Planning for the cache miss before the spike.
A potential Redis expiry stampede was addressed with simple proactive cache warming rather than distributed locking.
Simple · predictableRead →
Development workflow
AI-assisted development, with engineering judgement still in charge.
CLI-based AI tools are used as part of the development loop for coding, exploration, refactoring, debugging, documentation and code review. Depending on the task, that includes Claude Sonnet, Haiku or Opus, ChatGPT, and open models such as GLM and Kimi. Architecture, trade-offs and final code remain human decisions.
Current scope
Digital payments · cashback · lending · NPS · referral programmes · banking integrations