Best-in-Class Event Streaming
for Agents, Apps, and Analytics
for Agents, Apps, and Analytics
Sign up for our to stay updated.
Customer Story
Atome replaced their Apache Flink pipeline with RisingWave for real-time risk management. Feature delivery dropped from weeks to approximately one day.
The Challenge
Atome's risk management team relied on Apache Flink for real-time feature engineering, but the platform required specialized Flink expertise that was difficult to hire for. Each new feature took weeks to develop, package, and deploy, and OLTP query costs spiked during peak transaction periods.
Complex packaging and deployment pipelines slowed every update. Iterating quickly on risk scoring models was nearly impossible, creating a bottleneck that held back Atome's buy-now-pay-later fraud prevention capabilities.
The Solution
RisingWave replaced Flink operators with PostgreSQL-compatible SQL queries. Cascading materialized views continuously compute risk features, fed by MySQL CDC and Kafka sources. The familiar SQL interface eliminated the need for specialized streaming engineers and dramatically shortened iteration cycles.
Any SQL-proficient engineer can now modify risk pipelines directly. Rule deployment was reduced from weeks to 1-3 days, enabling Atome to respond to emerging fraud patterns in near real time.
Results
Atome reduced feature delivery from weeks to approximately one day and rule deployment to 1-3 days. The system processes 10,000 rows per second at the source and 2,000 rows per second at the sink, with sub-second end-to-end latency and zero backpressure events.
| Metric | Before | After (RisingWave) |
|---|---|---|
| Feature delivery | Weeks | ~1 day |
| Rule deployment | Weeks | 1-3 days |
| Source throughput | Limited | 10,000 rows/s |
| Sink throughput | Limited | 2,000 rows/s |
| End-to-end latency | Seconds+ | Sub-second |
| Backpressure events | Frequent | Zero |
Replace Flink complexity with SQL simplicity.
Accelerate Your Feature Delivery →