Explore the intricacies of real-time machine learning in this 52-minute podcast episode featuring Sasha Ovsankin and Rupesh Gupta, tech leaders from LinkedIn. Delve into the challenges and benefits of transitioning from batch to near real-time ML, covering topics such as real-time inference, feature engineering, and the tools and steps required for implementation. Gain insights on when to make this transition, the importance of feedback loops, and the technical details behind integrating recent information into models. Learn about embedding vector search, the true meaning of "real-time" in ML contexts, and the scenarios where real-time features are most beneficial. Discover the successes and learnings from implementing real-time features at LinkedIn, and understand the often-overlooked aspects of real-time ML. This comprehensive discussion provides valuable knowledge for ML practitioners looking to enhance their systems' responsiveness and effectiveness.
Real-Time Machine Learning: Features and Inference - MLOps Podcast #135
Overview
Syllabus
[] Sasha's and Rupesh's preferred coffee
[] Takeaways
[] Changes in LinkedIn
[] "Real-time" Machine Learning in LibnkedIn
[] Value of Feedback
[] Technical details behind getting the most recent information integrated into the models
[] Embedding Vector Search action occurrence
[] Meaning of "Real-time" Features and Inference
[] Are "Real-time" Features always worth that effort and always helpful?
[] Importance of model application
[] Challenges in "Real-time" Features
[] System design review on Pinterest
[] Successes of real-time features
[] Learnings to share
[] Branching for Machine Learning
[] Not so talked about discussion of "Real-time"
[] Wrap up
Taught by
MLOps.community
Related Courses
-
Machine Learning for Musicians and Artists
4.8 -
Real-time Machine Learning - MLOps Coffee Sessions
-
RecSys at Spotify - Building and Evaluating Large-Scale Recommender Systems
-
MLOps vs ML Orchestration - Exploring Challenges and Community-Driven Development
-
Building Real-Time ML Pipelines the Easy Way
-
Need for Speed: Machine Learning in the Era of Real-Time
