Twinder Microservice
Event-driven backend designed for traffic spikes and consistent latency across a social matching workload.
View source on GitHub ↗
Role
Backend Engineer
Highlights
- Scale: Java microservices processed about 500M daily requests with Redis caching and AWS EC2 autoscaling for sustained performance.
- Streaming: Kafka batch compression and partitioning fanned out workloads to consumers, cutting dispatch time from 200ms to 100ms.
- CQRS + Sharding: CQRS storage design accelerated reads and writes while a sharded MongoDB cluster supported 99.98% uptime under peak load.
- Performance testing: Multi-threaded clients and JMeter drove 2M requests, improving throughput from 2k to 6.6k req/s and latency from 200ms to 30ms.
Architecture
The system ingests swipe events at high volume, emits them to Kafka topics, updates read projections, and pushes match candidates into Redis. RabbitMQ supports targeted fan-out to user-specific queues for real-time updates.
Notes
Reach out if you want to walk through design choices, bottlenecks, and tradeoffs in more detail.