Brad Burega

Software Engineer

Spade

Roadblock Title:

Real-Time ML in 20ms: Building Low-Latency Models for Financial Decisions

Time:

Tuesday - 10:30 AM (Tower C)

Abstract:

At Spade, our core product is transaction enrichment - matching a credit card transaction to a counterparty with ultra-low latency. We’re continuously seeking to improve our match rate and our match accuracy, but in the real world, some matches are better than others. We use a confidence score to communicate to customers the quality of each match. However, when your API’s p50 is 23ms, that doesn’t leave a lot of time for computation. In this talk, we’ll explore how we built a low-latency neural network using the JAX framework to score the quality of each transaction, allowing customers to make informed decisions in the most sensitive use cases.

Bio:

Brad Burega is a software engineer at Spade working on low-latency APIs for fintech infrastructure. Previously, he worked on search relevance at Yelp. He has an MSc in Computer Science focused on Reinforcement Learning from the University of Alberta and a BSc in Computer Science and Physics from the University of British Columbia.