No. 498 of 500 No. 82 in Hands-On AI & Programming

Building Machine Learning Pipelines

Automating Model Life Cycles with TensorFlow

By Hannes Hapke and Catherine Nelson

Print length
366 pages
Publisher
O'Reilly Media
Reading level
Intermediate
ISBN-13
978-1492053194
Published
August 18, 2020
Written for
Data scientists and ML engineers moving TensorFlow models from experimentation into automated production.

Synopsis

Hannes Hapke and Catherine Nelson address the gap between training a model in a notebook and running it reliably in production. Their answer is an automated pipeline, and the book builds one step by step with TensorFlow Extended (TFX), Google's open-source framework for production machine learning. A running example carries the reader through data ingestion, data validation with TensorFlow Data Validation, feature preprocessing with TensorFlow Transform, model training, and model analysis with TensorFlow Model Analysis, including checks for fairness across subgroups. Later chapters cover deployment with TensorFlow Serving and other options, privacy-preserving techniques, custom TFX components and feedback loops that retrain models as new data arrives.

Orchestration receives detailed treatment, with chapters showing how to run the same pipeline on Apache Beam, Apache Airflow and Kubeflow Pipelines. The authors write for data scientists and machine learning engineers, but they also explain to managers why automation shortens the path from experiment to deployment and makes models auditable. Readers gain a concrete mental model of the stages every production system needs, from validation to monitoring, plus working code for each stage. The TFX APIs shown date from 2020, so readers today will need to check current documentation, and teams built around PyTorch or newer MLOps platforms will apply the concepts more than the code.

Scorecard

Editorial score

How it scored

Editorial score: 62.3 / 1005 weighted criteria

  • Influence30%

    18.0/30
  • Authority25%

    17.5/25
  • Practical value20%

    13.0/20
  • Currency15%

    6.8/15
  • Readability10%

    7.0/10
  • Editorial score

    62.3/100

How the rank was computed

Rank No. 498 of 500 comes from a composite of 46.2: observable signals 18.1 of 55 points plus editorial review 28.1 of 45 points.

All 11 inputs
InputPoints
Observable signals18.1/55
Reach (Amazon ratings volume)6.8/16.5
Consensus across best-of lists5.5/13.8
Awards and bestseller evidence0.0/8.3
Reader rating, sample-adjusted5.6/8.3
Amazon sales rank0.2/5.5
Recency0.0/2.8
Editorial review28.1/45
Author authority7.9/11.3
Influence8.1/13.5
Practical value5.9/9.0
Readability3.2/4.5
Currency3.0/6.8

Each criterion is scored from 1 to 10 and weighted, and the points add up to the editorial score. The rank combines observable signals (55 percent) with this score (45 percent). Read the methodology.

Verdict

A clear, end-to-end walkthrough of production ML pipelines on TensorFlow Extended; its principles endure, but its tooling has aged.

Strengths

  • Complete worked example from data to deployment
  • Covers fairness and privacy alongside engineering
  • Compares three pipeline orchestrators in depth

Keep in mind

Tightly bound to TensorFlow and 2020-era TFX releases, so PyTorch users and teams on newer MLOps platforms will find much of the code dated.

Key takeaways

  1. Automating the full model life cycle, from ingestion to deployment, saves engineering time and makes every model version reproducible and auditable.
  2. Validating incoming data against a learned schema catches drift and anomalies before they silently degrade a model in production.
  3. Model analysis should slice performance by user subgroup, exposing fairness problems that aggregate accuracy metrics conceal.
  4. Pipeline logic can stay independent of the orchestrator, letting teams move between Apache Beam, Airflow and Kubeflow Pipelines as needs change.

Notable facts

  • Co-author Hannes Hapke is community lead and moderator of Google's TFX Special Interest Group and later co-wrote Machine Learning Production Systems (O'Reilly, 2024). Source (opens in a new tab)

About the author

Hannes Hapke and Catherine Nelson

Hannes Hapke is an ML engineering leader who heads Dataiku's 575 Lab open-source office and was the first ML engineer at Digits; he moderates Google's TFX Special Interest Group and co-wrote Machine Learning Production Systems and Natural Language Processing in Action. Catherine Nelson, a former principal data scientist at SAP Concur with a PhD in geophysics from Durham University, wrote Software Engineering for Data Scientists.

Questions

Is Building Machine Learning Pipelines still worth reading in 2026?
Its concepts remain sound: data validation, reproducible preprocessing, sliced model evaluation, automated deployment and feedback loops are still the backbone of production machine learning. The code, however, targets 2020-era TensorFlow Extended releases, so readers should expect API changes. Teams standardized on TensorFlow will benefit most; others should read it for architecture rather than implementation.
Who should read Building Machine Learning Pipelines?
Data scientists, machine learning engineers and DevOps engineers responsible for getting models into production will gain the most, particularly in organizations that use TensorFlow. Readers should know Python and basic model training. Managers can read the opening and orchestration chapters to understand what an automated pipeline involves and why it shortens deployment cycles.
What does Building Machine Learning Pipelines cover?
The book builds a complete pipeline with TensorFlow Extended, covering data ingestion, validation, preprocessing with TensorFlow Transform, training, model analysis including fairness checks, deployment with TensorFlow Serving, privacy techniques, custom components and feedback loops. It then shows how to orchestrate the pipeline with Apache Beam, Apache Airflow and Kubeflow Pipelines, so teams can run it on their preferred infrastructure.

Sources

As an Amazon Associate, Best AI Books earns from qualifying purchases. Best AI Books is sponsored by Iternal.ai, whose founder and CEO, John Byron Hanby IV, wrote The AI Strategy Blueprint, our #1 ranked book. How we rank books.