Case Study: Qraft Technologies boosts AI-powered order execution performance with StreamNative's Apache Pulsar

Qraft Technologies Increases AI-Powered Order Execution Performance with Apache Pulsar

Qraft Technologies, a financial services company, needed an event streaming platform for its AI-powered order execution system that could handle massive South Korean market tick data with high throughput, low latency, and high availability. After evaluating Apache Kafka and finding its latency too high for heavily scaled topics, Qraft turned to StreamNative’s Apache Pulsar offering to support its real-time machine learning execution model.

StreamNative helped Qraft implement Apache Pulsar across the AXE architecture, including live tick-data ingestion, job distribution, and microservice communication. After about a month of testing and tuning, Qraft achieved end-to-end latency below 10 ms and then below 5 ms, with a future goal of under 1 ms. Pulsar also passed failover tests, and Qraft was able to build a custom Rust client and contribute it back to the community.

Qraft Technologies