moco
30-50% fewer FLOPS in inference -> lower latency, higher throughput, fewer missed SLAs.
Problem:
Fraud Detection models along the critical path must return responses within critical time windows (called SLAs) or be penalized. In moments of high throughput, transactions get queued and SLAs get missed.
For Citibank, a company with 10M transactions / day, assuming 0.1% missed SLAs, assuming $10 / missed SLA penalty that is $100k/day.
Solution:
moco offers a drop-in solution: a mathematical optimization library that takes in a trained model and a dataset and outputs a more efficient model, saving computations, latency, less queuing and thus fewer missed SLAs.
Key Benefits:
Interested? Schedule a demo today!
Contact:
Sam Randall, Founder