Explore the MATS stack for cross-system orchestration of machine learning pipelines in this 22-minute conference talk from Databricks. Learn how Avast integrates model tracking, storage, orchestration, and deployments to handle over 17 million daily phishing detections. Discover how MLFlow, Airflow, Tensorflow, and Spark combine to create a standardized, well-integrated toolset for data scientists. Follow the journey of Angler, an internal project for detecting phishing URLs, through all pipeline stages including data transformations, model training, experiment tracking, and serving. Gain insights into fast, reproducible experiments and seamless progression from research to production. Understand the challenges and successes of implementing this modern ML pipeline approach, which can be integrated into existing ecosystems without disruption.
Overview
Syllabus
Introduction
Project Life Cycle
MATS Stack
Airflow
Tensorflow
Challenges
Successes
Taught by
Databricks
Related Courses
-
MLOps Course – Build Machine Learning Production Grade Projects
-
MLOps Platform Architecture for End-to-End ML Pipelines
-
Productionizing Machine Learning Pipelines with Databricks and Azure ML
-
MLflow Model Serving - From Training to Production
-
Quick to Production: Integrating Spark and TensorFlow for Efficient MLOps
-
Introducing MLflow for End-to-End Machine Learning on Databricks
