Lightning Talk: SLIs for ML Models
Presenter: Sunandan Barman, Production Engineer, Meta Platforms
Abstract:
As machine learning (ML) models become increasingly critical to business operations, ensuring their reliability and performance is paramount. This talk explores the challenges of implementing Service Level Objectives (SLOs) for ML models and presents best practices for defining and monitoring custom SLOs that account for ML model specifics.
Using a case study, we demonstrate how implementing SLOs can lead to significant improvements in model accuracy and inference latency. Attendees will learn how to create effective SLOs for their ML models and improve the overall reliability and performance of their machine learning services.
By attending this talk, you will gain a clear understanding of how to ensure the reliability and performance of your ML models, and how to define and monitor SLOs to achieve this goal.
Presenter Bio:
Sunandan Barman, Production Engineer, Meta Platforms
Sunandan Barman is a highly experienced backend developer with more than a decade of expertise in crafting scalable and distributed systems capable of handling immense loads. He has delivered large-scale projects consistently is proficient in setting direction for cross-functional teams on multi-year hyper-growth paths.
Linkedin : https://www.linkedin.com/in/sunandanbarman/