How to aviod suffereing in MLOps

And I did it again, if you got a broken link in the last email while trying to get to the second session of the engineering labs recap, my bad. here is a good link.

We now have weekly office hours which are a way for us all to gather and have informal chats that are not recorded. We meet Thursdays at 8am PST/ 4pm GMT. If you would like to be added to the invite let me know.

We are going to be talking to the famous D. Sculley whose papers you have probably read at some point or another. we wanted to source some questions from everyone since he doesn't do many public appearances. So let us know, did you have any questions you wanted to ask the author of Machine Learning: The High Interest Credit Card of Technical Debt?

Call For Participation

The Past: As you are probably well aware by now we have finished our first MLOps Engineering Labs with outstanding results. Apart from the coffee sessions we recorded with team 1 and team 3, we also now have the technical deep dive write up with code snippets and all! Also check out the github repo's of team 1 and team 3 for more goodness.

The Present: As of now we are still allowing submissions for the next voyage which will begin early April. We will do the big reveal the week before things get rolling. Click the button below and get involved!

Product Management In ML

Big shout out to the most prolific man on our MLOps community slack @Laszlo Sragner for his participation in another epic meetup last week. Considering the traction and chatter we had during the meetup I feel like I wasn't the only one enjoying his perspective.

What did we learn? There are so many nuances when it comes to managing an ML project and they are worlds apart from normal software deliver cycles... or are they? I really like that Laszlo says get an end-to-end MVP as soon as possible which usually entails that the model is simple.

Greatest Takeaway Ever: To break with the Waterfall process you need an organisational structure that can attempt each part of the model lifecycle at the same time. This requires a rigorous version control of data and code what we call "Chain-of-Evidence" and ensemble modelling. You need a well organised MLOps infra so you can focus on the above rather than struggle with low level issues.

+ It's not so easy to find PMs who have had experience in ML before, which begs the question, is it a prerequisite? Laszlo summed it up nicely, and another answer came from Korri in the chat "I think there is a place for product management for those that have been engineers/data scientists. However, there is the capability of managing an ML product, but then there is the case of having a "VISION", which is an existing gap across"

+All in all it was a 10/10 meetup. Check out the video here and the podcast here.

We're Getting A Facelift

If I had a dollar every time someone told me the website was out of date or the slack link didnt work I'd be a rich man.

The Time Has Come: Yes my friends, we have finally started to give the website a much-needed makeover and because of that, we need your help!

The Premise: On the new website you will be able to see everything that you would expect, a gallery of all the past meetups and coffee sessions, a link to slack or to sign up for the newsletter, our own blog, AND something really special; a tooling comparisons guide. As of now we are releasing these three capabilities charts and asking for your feedback. One for ML Monitoring, Feature Stores and one for Model Deployment. What are some capabilities that we missed?

We want this project to be a community-sourced effort. We have already talked to many of the top companies in each space to get their take on these lists but we don't want it to stop there, it should be a community-wide effort! Please add anything else you think is important as a comment in the doc and we will incorporate it into the framework. Oh yeah and someone call Gartner and tell them we are doing their job for them.

Suffering Is Inevitable

This week we will be talking with none other than Igor Lushchyk a data engineer at Adyen who for the last 5 years has had his head down doing MLOps work.

Meetup: The theme of our conversation is 'How to Avoid Suffering in MLOps' lessons learned from 5 years of hitting your head against the wall. We plan to cover the transition from homegrown platform to open source solutions, supporting old solutions and maturing them all the while making data scientists happy. Now you see why Buddha was wise?

Bio: Igor is a member of the core data infrastructure team developing the model productization pipeline, data experimentation, and ETL platforms. He has architected or contributed to multiple parts of given systems, driving projects end-to-end, from stakeholder communications and specs to integration.

+ As always see you at 5pm GMT / 9am PST tomorrow, Wednesday by clicking the link below.