Liquid AI
5 Case Studies
A Liquid AI Case Study
A top tier global payment processor faced the challenge of identifying fraud in real-time across billions of daily transactions. Key obstacles included processing up to 65,000 transactions per second with sub-2ms latency, long model training cycles, and massive state storage constraints that risked petabytes of additional data. Liquid AI partnered with the client to address these issues.
Liquid AI developed a custom model powered by its Liquid Foundation Model backbone. The solution delivered a compact state representation for scalable storage, a CPU-optimized inference stack for sub-1ms latency, and efficient training techniques. The results included an estimated $230 million in additional fraud losses prevented annually, over 2x faster prediction latency, and 1.8x faster training, all while being deployed on-premise for privacy compliance.
Top Tier Global Payment Processor