Abstract: The benefits of Real-Time Machine Learning are becoming increasingly apparent. Digital native companies have long proven that use cases like fraud detection, recommendation systems, and dynamic pricing all benefit from lower latencies. In a recent KDD paper*, Booking.com found that even a 30% increase in model serving latency caused a .5% decrease in user conversion, a significant cost to their business.
While Real-Time Machine Learning presents a game-changing opportunity, few data teams are doing real-time serving in production, struggling to deliver the feature freshness needed for low-latency inference.
Real-Time Machine Learning has yet to reach its potential because of the deep disconnect between data engineering and data science. Historically, our industry has perceived streaming as a complex technology reserved for experienced data engineers with a deep understanding of incremental ingestion patterns. But now, modern streaming platforms make it much easier for anyone to build reliable streaming pipelines, regardless of their streaming background.
Using an example fraud detection scenario, you’ll learn:
- Three important patterns for real-time model inference
- How to prioritize the most common real-time ML use cases in your business
- How to evaluate streaming tools, and why streaming is valuable at any latency
- Operational concerns like monitoring, drift detection, and feature stores
Who should attend: data scientists regardless of streaming experience, and data engineers regardless of ML experience.
Bio: Dillon Bostwick is a Solutions Architect at Databricks, where he’s spent the last five years advising customers ranging from startups to Fortune 500 enterprises. He currently helps lead a team of field ambassadors for streaming products and is interested in improving industry awareness of effective streaming patterns for data integration and production machine learning. He previously worked as a product engineer in infrastructure automation.