My MLOps bookshelf
Over the past three years, the topics on my bookshelf have changed to lean more towards MLOps; I have read too many books on the subject that I can finally choose my favourites; let’s have a look:
My favourites
Building Machine Learning Powered Applications
This book by Emmanuel Ameisen (O’Reilly) was the one that pushed me to leave my job as a data scientist turned data engineer and start moving my career towards MLOps.
The book is very lightweight in practical content; what I loved about it is the high-level view that takes away the focus from just developing the model and puts it into the surrounding components of a complete machine learning system. Buy from Amazon: Worldwide / México.
Reliable Machine Learning
This book by Cathy Chen et al. (O'Reilly) was a really nice find. It is technology agnostic, and despite not mentioning MLOps that much, MLOps is the subject of the entire book.
It puts ML through a lens of the concept of reliability, covering a wide range of topics, from issues that can happen at the data and storage layer, to glossing over some techniques to calculate costs, to who owns model quality. Buy from Amazon: Worldwide / México.
Machine Learning Engineering in Action
This is a densely packed book by Ben Wilson (Manning), worth every one of its more than 500 pages that cover all aspects involved in putting a successful machine learning application in production, including the planning and scoping of a project, experimentation, testing, deployment and monitoring. The whole cake!
This book is full of examples and diagrams that make it easy for the concepts to land. Highly recommended. Buy from Amazon: Worldwide / México.
Effective Data Science Infrastructure
What I love about this book by Ville Tuulos (Manning) is that, despite being a technical book centred around a tool –Metaflow–, the concepts it touches are timeless when it comes to what a good experimentation and deployment platform looks like.
To our benefit, the author draws from their experience building the data science infrastructure for one of the big tech companies. Even if you are not interested in Metaflow at all, I highly recommend checking this book. In my case, this book was a source of inspiration when designing the data science experience at my current company. Buy from Amazon: Worldwide / México.
Machine Learning Design Patterns
While this book by Valliappa Lakshmanan (O’Reilly) is not about MLOps as such, I still found it an interesting reference book that helps solving some of the common problems an ML-related engineer can find on a day-to-day basis.
Don’t think of this book as the one that will guide you from start to finish while building a machine learning app. Think of this book as a recipe cookbook you can keep referencing to. There is some criticism regarding the technology choices of the authors, but to me, the value shared in this book outweighs the annoying push towards “promoting” GCP, Tensorflow and BigQuery. Buy from Amazon: Worldwide / México.
Designing Machine Learning Systems*
I am still making my way through this book by Chip Huyen (O’Reilly), but so far it is promising — hence its position on the list. Will update this listing when I finish it. Buy from Amazon: Worldwide / México.
These books are good
Machine Learning Engineering with Python
A book by Andrew P. McMahon (Packt) that tries to cover many topics but in its short form barely gives any depth to any of it. What is good is that it covers some of the aspects that other books do not, such as Git workflows and Python packaging.
It also has some self-assessment questions that are good to test your knowledge as you go through the book. Buy from Amazon: Worldwide / México.
MLOps engineering at Scale
A relatively small book by Carl Osipov (Manning), it is a lightweight introduction to PyTorch and how to serve it in AWS. It has good ideas but I cannot get rid of the feeling that its size gets in the way of deep explanation. If you work with the technologies the book covers, it would be a good addition to your library. Buy from Amazon: Worldwide / México.
Machine Learning Engineering
A good reference book by Andriy Burkov (True Positive Inc) is a concise, albeit unorganised, overview of the machine learning engineering field. It is not a book that will greatly change your perspective on machine learning in production but it has some good valuable ideas.
I recommend this to that data scientist that is just beginning and wants to communicate with experienced data scientists or machine learning engineers, but if you already have experience, it won’t be super useful. Buy from Amazon: Worldwide / México.
Introducing MLOps
As the title suggests, this book by Mark Treveil et al. (O’Reilly) is a good enough introduction to MLOps; I’d recommend this book if you are entirely new to the subject. Go in without expecting too much technical depth and more of a high-level (enterprise?) overview. Buy from Amazon: Worldwide / México.
Recommended only if…
Building Machine Learning Pipelines
Book by Hannes Hapke et al. (O’Reilly), is just a cookbook that resembles the Tensorflow Extended documentation, except a bit outdated.
I don’t give a worse opinion of this book because the authors were upfront about using Tensorflow on the cover; I knew what I was getting into. If Tensorflow is your technology of choice, go for it — though, be aware of the changes in the TFX API since the book was released! Buy from Amazon: Worldwide / México.
Engineering MLOps
Sadly, this book by Emmanuel Raj (Packt) is too shallow and tries to cover too much ground but fails to land any particular topic with confidence.
The author uses Azure to build the project referenced throughout the book (when I bought the book this was a surprise to me as it was never referenced anywhere in the book description). You may like this book if you already know about MLOps and need to work with Azure as a cloud backend. Otherwise, I can’t recommend it. Buy from Amazon: Worldwide / México.
Machine Learning Systems
If you use Scala for your day-to-day, consider this book by Jeff Smith (Manning). If you are not into Scala, skip it. The book is too shallow for my liking, and talks more about software development with Scala than machine learning (or their relationship to ML). Buy from Amazon: Worldwide / México.
I can’t recommend these
Practical MLOps
I can’t make out what this book by Noah Gift et al. (O’Reilly) is about, but definitely not so much about MLOps. There are loads of irrelevant examples and personal experiences that feel forced. But the killer for me are the neck-breaking changes of topics from chapter to chapter; it reminded me of when I had to do teamwork with people I did not get along with back in high school.
Probably one of the worst O’Reilly books I’ve read in my tech career. Steer clear of it.
Conclusion
I hope this list gives you a good idea of what books to add to your bookshelf (or which ones to avoid), and if you have already read the ones I mention, feel free to comment on them and even politely disagree with me in the comments.
I am always open to book suggestions, so drop those in too in the comments.
