Artificial Intelligence #10: An easy way to explain MLOps – CI + CD + CT

Welcome to edition #10

We crossed 12K subscribers in a little over 8 weeks

Thanks for your support

This week, I will briefly cover MLOps

While MLOps is in the limelight now, MLOps has always been a key part of our course at the #universityofoxford (we open admissions for fall next week for #artificialintelligence: cloud and edge implementations course

I was asked to explain MLOps in a simple way at a session last week at the Caledonian club . Sean Connery was a member; So was John Logie Baird (inventor of the television) and Sir Alex Ferguson is a member – I was an invited guest ??

Quite a sudden change for a post lockdown dress code!

So, here is my explanation – I hope you find it useful also

The easiest way to explain MLOps is as CI + CD + CT

To elaborate

• The term?DevOps?comes from the software engineering world and is concerned with developing and operating large-scale software systems.?DevOps introduces two concepts:?Continuous Integration (CI) and Continuous Delivery (CD).?
• Continuous integration (CI) is the practice of automating the integration of code changes from multiple contributors into a single software project.(Atlassian)
• Continuous delivery is an approach where teams release quality products frequently and predictably from source code repository to production in an automated fashion.(Atlassian)
• DevOps aims to shorten development cycles, increase deployment velocity and create dependable releases by combining CI and CD
• The term?MLOps?refers to a combination of machine learning and devops. MLOps is a set of techniques and practises for data scientists to collaborate operations professionals.. MLOps aims to manage deployment of machine learning and deep learning models in large-scale production environments.
• Since, an ML system is also a software system, DevOps principles also apply to MLOps.

?The overall steps for deploying an ML/DL model in production are:

• Data extraction
• Data analysis
• Data preparation
• Model training
• Model evaluation
• Model serving
• Model monitoring

?While MLOps encompasses DevOps, there is a significant difference between the two

ML and DL systems are impacted by changing data profiles. This is not the case in a traditional IT system. Hence, the model has to be refreshed even if it ‘works’ currently – leading to more iterations in the pipeline. Hence, you have to monitor models in production and refresh the model by retraining if the model performance falls below a certain criteria (continuous training).

?Hence, we can think of MLPOps as CI + CD (traditional DevOps) + CT (continuous training)

?

I explain this using the diagram below (from Azure). The red boxes show the CI CD and the CT segments

?

Trust that was useful for you also

?In terms of jobs

?Aimii is hiring machine learning interns? ?

Alberta Machine Intelligence Institute (Amii) ?As one of Canada’s preeminent centres of artificial intelligence and ?one of Canada’s three centres of AI excellence. Their team includes Prof Matt Taylor (who is also part of our Oxford course) and also Richard Sutton (of Sutton and Barto ). Through Matt, we also know Cam Linke and Stephanie Husby of Aimii. So, I have no hesitation in recommending this role

Finally, many thanks to Paul Lin knewtopia in china who are backed by Tongji University's venture accelerator for the session at the Caledonian club. Also thanks to the very professional team at Cambridge film works for the video production.

Image source: Azure