Practical MLOps Part 2 // Alfredo Deza // MLOps Meetup #66

MLOps.community

MLOps community meetup #66! Last Wednesday we talked to Alfredo Deza, Author and Speaker.
//Abstract
In this episode, the MLOps community talks about the importance of bringing DevOps principles and discipline into Machine Learning. Alfredo explains insights around creating the MLOps role, automation, constant feedback loops, and the number one objective - to ship Machine Learning models into production.
Additionally, we covered some aspects of getting started with Machine Learning that is critical, in particular how democratization ML knowledge is critical to a better environment, from libraries to courses, to production results. Spreading the knowledge is key!
//Bio
Alfredo Deza is a passionate software engineer, speaker, author, and former Olympic athlete. With almost two decades of DevOps and software engineering experience, he teaches Machine Learning Engineering and gives lectures around the world about software development, personal development, and professional sports.
Alfredo has written several books about DevOps and Python including Python For DevOps and Practical MLOps. He continues to share his knowledge about resilient infrastructure, testing, and robust development practices in courses, books, and presentations.
Alfredo Deza is the author of Python for DevOps and Practical MLOps.
----------- Connect With Us ✌️-------------
Join our Slack community: https://go.mlops.community/slack
Follow us on Twitter: @mlopscommunity
Sign up for the next meetup: https://go.mlops.community/register
Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Connect with Chris on LinkedIn: https://www.linkedin.com/in/chrisbergh/
Timestamps:
[00:00] Introduction to Alfredo Deza
[03:00] Alfredo's background in tech
[13:15] Who is this book for?
[14:15] "The reason why we need a Machine Learning book is that there's definitely a knowledge gap."
[16:05] Hierarchy of MLOps
[17:16] "Automation has to be the basis of pretty much, everything."
[19:03] Logging - "When in doubt, log it out!"
[24:50] Maturity
[29:55] "The notion of self-healing is very appealing."
[31:20] Learning Test
[37:40] "Catch things as early as possible. Anything that comes at the end of the process, the closer you are to the production, the more expensive it could get."
[37:54] "Expensive can be the dollar amount in engineering time, or it can be the dollar amount in services that you're using to produce, and the dollar amount on how long it would take to ship the version that fixes the problem."
[39:20] "Why not scan your containers before they hit the production and catch anything that has a critical vulnerability announced?"
[40:08] Interrupibility standards and pains
[42:34] "It is critical that we make it easier. How about we no longer point fingers and stigmatize people who don't do Machine Learning. The more people doing Machine Learning today, the better we're off."
[45:50] Simple and opinionated or flexible and complex
[46:45] "You have to strike a balance but you have to stay true to your principles."
[50:38] Abstraction Layers
[56:57] Take a risk or stay safe?
[57:20] "I think, you're gonna have risk everywhere you are. You're gonna have risk when you hire a Machine Learning Engineer. You're gonna have a risk with a Data Scientist. You're gonna have a risk with a Software Engineer."