Hi ,

The MLOps reading group that was announced in the last newsletter now has a date set for the first meetup! Next Friday, March 19 at 5pm UTC, we'll be discussing the classic Google article "Rules of Machine Learning". If you're interested in participating, join the # reading-group channel in the MLOps Community Slack and subscribe to the update email list here.

In this week's edition:

• Bringing DevOps Best Practices into Machine Learning
• Developing Feast, the Leading Open Source Feature Store
• What are feature stores, anyway?
• Facebook Advances SoTA in Computer Vision Self-Supervised Learning
• Practical Advice for "Managing Up"

Bringing DevOps Best Practices into Machine Learning

Benedikt Koller is a self-professed "Ops guy", having spent over 12 years working in roles such as DevOps engineer, platform engineer, and infrastructure tech lead at companies like Stylight and Talentry in addition to his own consultancy KEMB. He's recently dove head first into the world of ML, where he hopes to bring his extensive ops knowledge into the field as the co-founder of Maiot, the company behind ZenML, an open source MLOps framework.

In this episode, Benedikt discusses common problems faced by teams putting machine learning into production, bringing over best practices from DevOps to solve them, and building ZenML, an open source MLOps framework.

Click here to listen to the episode

Developing Feast, the Leading Open Source Feature Store

Willem Pienaar is the co-creator of Feast, the leading open source feature store, which he leads the development of as a tech lead at Tecton. Previously, he led the ML platform team at Gojek, a super-app in Southeast Asia.

In this episode, Willem discusses his experience building the ML platform at Gojek and what he's learned from developing and open-sourcing Feast. He also goes into his vision for it's future and how teams can best get started adopting it.

Click here to listen to the episode

What are feature stores, anyway?

Before listening to the episode above, if you're not already familiar with the concept of a feature store, I highly suggest reading this article that Willem co-authored. It covers the problems a feature store solves, where it integrates into your system, and some of the core elements that make up a feature store.

For more on the topic, also see Eugene Yan's recent post "Feature Stores - A Hierarchy of Needs"

Facebook Advances SoTA in Computer Vision Self-Supervised Learning

"We believe that self-supervised learning (SSL) is one of the most promising ways to build such background knowledge and approximate a form of common sense in AI systems."

Ever since interviewing Josh Albrecht, who's done extensive work on SSL recently, I've been fascinated by the progress that's been happening in this sub-field.

Last week, Facebook released their paper and code for SEER, a 1.3B-parameter model trained on uncurated, unlabeled, real-world data that was able to achieve state-of-the-art results on ImageNet. In addition, they were also able to achieve good accuracy in a few-shot setting.

Facebook also published a blog post co-authored by Yann LeCun calling SSL the "dark matter of intelligence". It covers the challenges computer vision presents when applying techniques previously used in NLP and three of the approaches they find most promising: energy-based models, joint embeddings, and latent-variable architectures.

If you want a technical primer on self-supervised learning, check out Lilian Weng's excellent blog post here.

Practical Advice for "Managing Up"

"Managing up is usually a lot more practical. Your manager doesn’t (and can’t!) know every single detail about what you do in your job, and being aware of what they might not know and giving them the information they need to do their job well makes everyone’s job a lot easier."

While not related to the technical side of ML or software engineering, I found this blog post by Julia Evans to be incredibly useful. She details nine different things that you can communicate to your manager to help them help you and your team, with specific examples of each one.

Thanks for reading and have a great rest of your week!

Charlie