Data Science mit AWS: End-to-End-Pipelines für Continuous Machine Learning implementieren

by Chris Fregly

Mit diesem Buch lernen Machine-Learning- und KI-Praktiker, wie sie erfolgreich Data-Science-Projekte mit Amazon Web Services erstellen und in den produktiven Einsatz bringen. Es bietet einen detaillierten Einblick in den KI- und Machine-Learning-Stack von Amazon, der Data Science, Data Engineering und Anwendungsentwicklung vereint. Chris Fregly und Antje Barth beschreiben verständlich und umfassend, wie Sie das breite Spektrum an AWS-Tools nutzbringend für Ihre ML-Projekte einsetzen.
Der praxisorientierte Leitfaden zeigt Ihnen konkret, wie Sie ML-Pipelines in der Cloud erstellen und die Ergebnisse dann innerhalb von Minuten in Anwendungen integrieren. Sie erfahren, wie Sie alle Teilschritte eines Workflows zu einer wiederverwendbaren MLOps-Pipeline bündeln, und Sie lernen zahlreiche reale Use Cases zum Beispiel aus den Bereichen Natural Language Processing, Computer Vision oder Betrugserkennung kennen. Im gesamten Buch wird zudem erläutert, wie Sie Kosten senken und die Performance Ihrer Anwendungen optimieren können.

Genres: Artificial Intelligence, Technology, Technical

552 pages, Kindle Edition

Published April 13, 2022

Reviews

[Sebastian Gebski · Rating 3 out of 5]

It simply HAD to be a success: super-interesting topic (data science), amazing tech (SageMaker), the author who's a real expert on the topic, published who puts a lot of effort in keeping the bar high, ...

But it isn't (a success). I so wanted to fall in love with this book, but I didn't. Why?

1. The author really wanted to show as much Data Science goodness as possible. Unfortunately, as a result, the whole book feels like a roller-coaster or US tourists' trip across Europe ("10 capital cities in 2 weeks"). Zillion topics, but each of them covered in a rush, w/o setting proper foundations. Don't get me wrong, he does keep the proper structure - the chapters do have introductions, but after a good start (initial 2 pages of general context), he dives super deep into Python code ...

2. SageMaker is not a trivial service. There are some unobvious conceptual constructs that require careful elaboration. It's hard to navigate across such a complex service w/o a proper overview, possibly some examples that approach it from very different angles. Or a real-life study across the whole model's lifecycle. I think that's what is totally missing in official AWS docs and ... unfortunately, it's also what's missing in the book. YES, there are very specific code samples that I assume are correct and solve a narrow, particular problem. But does reading through them make the reader able to compose a solution to another (even similar) problem? I don't think so ...

3. I was wondering - who could be the best audience for this book. And I have a surprising conclusion: probably people who already know the SageMaker basics, who have used it in a single scenario or two. This book 'drafts' briefly so many scenarios that can spur their imagination (regarding what's possible in AWS when it comes to AI/ML) and they already have enough understanding of concepts behind the service to take it from here.

It really saddens me to rate this book so harshly, especially because it's very visible how much effort was put into it. But I can't recommend it to people who'd like to learn how to do Data Science on AWS :( Because they won't. I wouldn't be able to.

2.5-2.7 stars

[Xianshun Chen · Rating 3 out of 5]

Pretty good on the part such as how to use SageMaker Studio and AutoPilot as well as Athena

The later chapters on BERT and tensorflow is not easy to follow as the book is still in its early release. Also the SageMaker's ScriptProcessor to run Spark jobs need more effort to make the examples runnable.

[Mateusz · Rating 4 out of 5]

Very good book, but should be called Data Science with AWS SageMaker. There are examples of other tools/services, but these are minor. I get that SageMaker is the most developed and fitting solution in whole AWS Services Zoo, but there are other means/design patterns to achieve similar goals- I'd prefer more holistic view on the topic.

[Wojtek Gawroński · Rating 4 out of 5]

I really liked this book!

It's very actionable, comprehensive, full of examples and insights about the AWS portfolio - with a huge emphasis on Amazon SageMaker. Very good position for anyone interested in learning this set of services in practice.

[Suat Atan · Rating 5 out of 5]

The most comprehensive and inspiring book for AWS.
Most sources dictate to show steps and readers loss themselves among instructions.
This book is different it shows the way, the user can find rest.

[Philippe · Rating 3 out of 5]

The book is great to discover the different services and tools for AI/ML in AWS. But the code in the GitHub repository is very messy, and doesn't work anymore, which goes against the goal of the book.

[Ivan Matoshchuk · Rating 3 out of 5]

a bit outdated, but provides fundamentals.

[Scythe Rowan · Rating 4 out of 5]

Interessantes Buch, jedoch ist mir unklar, warum es nicht als Hardcover erschienen ist. 50+€ wären mir für ein Taschenbuch zu viel, da bin ich froh, dass ich es mir aus der Bücherei geholt habe...

[Klaus-Michael Lux · Rating 3 out of 5]

Covers a lot of ground quickly, demonstrating how large organisations with a lot of staff heft might put SageMaker, the AWS offering for semi-managed Data Science workflows to good use, but does make it seem a little overwhelming to start in case you don't have a dedicated DevOps team and a compliance department checking in on the correct S3 bucket access policy. The authors crammed a lot of examples and workflows in, trying to showcase all the various offerings in the SageMaker ecosystem and lots of special cases, but seem to lose track a bit of bringing more general messages across. It could be helpful to reorganize some of the insights from this book around user stories, trying to prove how sub-components of the product can be of value even before a massive effort has been made to adapt to the entire ecosystem. I'll definitely re-consider this book though for reference when wondering how particular workflows can be solved in the AWS ecosystem when they come up.

[Ntombizakhona Mabaso · Rating 5 out of 5]

The praise for this book, including from Jeff Barr, is enough to spark excitement for diving into it, I just hope it lives up to the hype! (It did)

It’s an excellent resource for data scientists looking to run workloads on AWS, offering a step-by-step guide that integrates the appropriate and relevant AWS services and technologies.

Even though I’m not a data science fanatic, I can appreciate its value as a comprehensive reference guide for those in the field.

[Usfromdk]

I read this one mainly for work-related purposes. It is still too early for me to really say if content from this book will prove useful or not, and as this is a crucial factor in the decision-problem of how to evaluate the book I decided for the time being not to rate it.