Most data science projects fail. There are various reasons why, but one of the primary reasons is the challenge of deployment. One piece to the deployment puzzle is understanding how data engineers can effectively work with data scientists to monitor and iterate on model performance, which is why we developed this course: MLOps1 (Azure) - Deploying AI & ML Models in Production using Microsoft Azure Machine Learning.

This is the second of three courses in the Machine Learning Operations Program using Azure Machine Learning.

Data Science, AI, and Machine Learning projects can deliver an amazing return on investment. But, in practice, most projects that look great in the lab (and would work if implemented!) never see the light of day. They could save or make the organization millions of dollars but never make it all the way into production. What’s going on? It turns out that making decisions in a whole new way is a big challenge to implement--for many technical, business andhuman-naturereasons. After decades of experience though, our team has learned how to turn this around and actually get working models into production the great majority of the time. A key part of deployment is excellence in data engineering, and is why we developed this course: MLOps1 (Azure): Deploying AI & ML Models in Production using Microsoft Azure Machine Learning.

You will get hands on experience with topics like data pipelines, data and model “versioning”, model storage, data artifacts, and more.

Most importantly, by the end of this course, you will know...

What data engineers need to know to work effectively with data scientists

How to embed a predictive model in a pipeline that takes in data and outputs predictions automatically

How to moniter the model’s performance and follow best practices

Syllabus

Week 1: The Machine Learning Pipeline

AI Engineering Role

ML pipelin lifecycle

Week 2: The Model in the Pipeline

Case Study for the Course

Model Undeerstanding

Week 3: Monitoring Model Performance

Logging and Metric Selection

Model and Data Versioning

Week 4: Training Artifacts and Model Store

Institution: StatisticscomX

Subject: Computer Science

Level: Intermediate

Prerequisites:

Predictive Analytics: Basic Modeling Techniques

Participants should be comfortable working with Python in a cloud-based environment, and will gain maximum benefit if they have some familiarity with software development, including git, logging, testing, debugging, code optimization and security.

Language: English

Video Transcript: English

Associated programs:

Professional Certificate in Machine Learning Operations with Microsoft Azure (MLOps with Azure)

Associated skills: Microsoft Azure, Return On Investment, Data Science, Data Engineering, Forecasting, Software Versioning, Operations, Machine Learning, Artificial Intelligence

What you\'ll learn

What data engineers need to know in order to work effectively with data scientists

How to use a machine learning model to make predictions

How to embed that model in a pipeline that takes in data and outputs predictions automatically

How to measure the performance of the model and the pipeline, and how to log those metrics

How to follow best practices for “versioning” the model and the data

How to track and store model and data artifacts