MLOps Masterclass: Theory to DevOps to Cloud-native to AutoML

Learn to go from theory to DevOps to MLOps platforms in this MLOps Master Class. 00:00 Intro 01:18 Noah Gift Background 04:14 Why do we need MLOPs? 05:06 Where the data science industry is headed? 06:57 Without DevOps you don't have MLOps 08:46 Continuous delivery is enabled by the Cloud and IAC 10:03 DataOps is like the water hookup in your home 11:23 Platform Automation solves the complexity of the data science industry 15:06 MLOPs Feedback loop 16:33 Create Once, but Deploy Everywhere. Good Example is Google AutoML 18:16 MLOps isn't data centric or model centric there is no silver bullet 21:52 MLOps use cases: Autonomous Driving is a good example 23:00 How to invest in technology: Primary and Secondary and Research 25:50 AWS and Azure are the leaders in the cloud 27:39 Secondary considerations: Splunk, Snowflake, BigQuery, Iguazio, etc 29:00 Leverage learning platform and metacognition 30:00 Key certifications 32:00 NFSOps is using managed file systems to build new cloud-native workflows 34:00 Kubernetes is the new gold standard for many distributed systems 35:00 Sagemaker has many use cases 36:21 Azure ML Studio 37:21 Google Vertex AI 37:48 Iguazio MLRun 41:00 Current issues in distributed systems 45:00 Apple Create ML Demo 51:00 Databricks Spark Clusters 57:00 MLFlow 01:00:37 What is DevOps? 01:03:16 Creating a new Github repo 01:05 Developering with AWS Cloud9 01:20:26 Setup Github Actions 01:23:00 Walkthrough of Python MLOps cookbook example using a sklearn project 01:35:00 Pushing sklearn flask microservice to Amazon ECR 01:39:00 Setup AWS App Runner for MLOps Microservice inference 01:43:00 Setup Continuous Delivery of MLOps Microservice using AWS Code Build 02:06:00 Comparing MLOps Platforms Databricks, Sagemaker and MLRun 02:31:00 Deploying MLRun open source MLOps with Colab Notebook