A Guide to Building a Continuous MLOps Stack with GCP and Superwise

Explore a comprehensive guide to building a continuous MLOps stack in this 1 hour 20 minute conference talk from MLOps World: Machine Learning in Production. Dive into MLOps CI/CD pipeline automation using GCP, Superwise, and retraining/auto-resolution notebooks. Learn how to construct a continuous ML pipeline for training, deploying, and monitoring models in the first part of the workshop. Discover automations and production-first insights for continuous issue detection and resolution in the second part. Gain insights from speaker Itay Ben Haim, an ML Engineer at Superwise, as he covers topics such as the importance of continuous training, the complexities of MLOps, Google Envelopes, pipeline overview and explanation, model training and evaluation, prerequisites, component functions, Flask service accounts, predict instances, and Docker builds. Enhance your understanding of MLOps and learn practical techniques for implementing efficient machine learning pipelines in production environments.

Intro
About Superwise
Why Continuous Training
Why MLOps is Complex
Google Envelopes
Scope
Notebook Setup
Pipeline Overview
Pipeline Explanation
Pipeline Use Case
Train Model
Evaluation
Next Steps
Prerequisites
Component Functions
Flask
Service Account
Predict Instance
Docker Build