Further Reading: The Data Engineering Career

Sources are tagged Tier 1 (confident it exists, recommended without reservation) or Tier 2 (real and worth seeking, but confirm the current edition, version, or URL yourself).

The strongest warning in the book's bibliography belongs here. Career advice is unfalsifiable, personal, and heavily selected for survivorship — the people who write it are the ones it worked for. It also ages worse than anything technical, because it describes a labour market rather than a system.

Read less of it than you think, weight practice over reading, and treat every list below as one person's experience generalized further than it should be — including this chapter's.

The three that are worth reading properly

Will Larson, Staff Engineer (2021), and An Elegant Puzzle (2019). The best available writing on the senior-and-above individual-contributor track, and unusually honest about the fact that staff roles are scarce and shaped differently at different companies. §40.4's claims are Larson's, with data-specific examples, and his archetypes — tech lead, architect, solver, right hand — are a more useful taxonomy than any levelling ladder. Tier 1.
Camille Fournier, The Manager's Path (O'Reilly, 2017). Nominally about management, and the chapters on the senior engineer and the tech lead are the best description of §40.3's transition anywhere. Read it even if you never intend to manage — partly because it tells you what management actually is, which is the information you need to decline it well. Tier 1.
Tanya Reilly, The Staff Engineer's Path (O'Reilly, 2022). The practical companion to Larson. The material on "big picture" thinking and on writing as the primary staff output is §40.4, worked through concretely. Tier 1.

On levels and the promotion problem

Your own company's levelling rubric, read carefully and then set aside. §40.2 and Case Study 2: the written criteria and the actual bar are different documents, and the way to bridge them is to ask for an instance rather than to re-read the criterion.
Publicly published engineering ladders — several companies have opened theirs, and levels.fyi-adjacent collections aggregate them. Read three, and notice what they share: the progression from your work to your outcome to other people's outcomes (§40.2) appears in nearly all of them, phrased differently. Tier 2 — availability varies; the pattern does not.
Anything on the "manager versus IC" fork, written by someone who went one way and observed the other. The version to be skeptical of is any that presents management as a promotion, which it is not, and any that presents it as a failure, which it also is not.

On skill durability

Chapter 40's durable list, read as a bibliography. Kimball (1996), Lamport (1978), and — for reconciliation — any introductory accounting text. The point of §40.11 is that these have not needed revision, which is itself the argument for reading them rather than this year's blog posts.
Fred Brooks, The Mythical Man-Month (1975, anniversary edition 1995). Fifty years old and mostly still right, which is a demonstration of §40.11's thesis rather than merely an example of it. Tier 1.
Any retrospective on a dead technology written by someone who used it. The Hadoop retrospectives are the closest analogue to Case Study 1, and the useful ones are those that separate what they learned from what they used. Tier 2 — scattered across blogs; the good ones are rare and worth saving when you find one.

On burnout, toil, and leaving

The Google SRE book's chapter on eliminating toil, and its 50% rule. §40.7's first failure mode, with a number attached and an argument for why it is an engineering problem rather than a personal resilience one. Tier 1.
Christina Maslach's work on burnout, which is the actual research rather than the folklore. The finding most relevant to §40.7 is that burnout correlates with lack of control more strongly than with workload — which is "custody without authority" named by a psychologist forty years earlier. Tier 2 — an academic literature; one review article is enough.
Chapter 26 of this book. On-call design, and the observation that a rota that pages more than twice a week is a rota that will lose people.

On the specializations

For each track in §40.5, this book's own chapters: streaming (29), platform (24, 27, 28), analytics engineering (18–20, 30), ML infrastructure (32), governance and privacy (30, 31). Read the chapter before committing to the track — it is a cheaper trial than a job.
Job postings, read as market data rather than as opportunities. Count how many roles in your market mention each track over a month. §40.5's demand column is exactly this measurement, and yours will differ from the book's — which is the point of doing it yourself (Exercise 40.7).

On the thing that is not about careers at all

Chapters 23, 30, 34, 36, and 38 of this book. §40.13's claim is that the durable skill is knowing whether a number is right, and these five chapters are the machinery for it: assertions, ownership, layer boundaries, reconciliation independence, and the capstone.
And the accounting literature on reconciliation, one more time. It is the oldest idea in this book (1494), it is the highest-weighted signal in §40.9, and it is the thing this whole chapter argues will still be your job when every tool named in these forty chapters has been replaced.

Practice

Start the running document. Exercise 40.1. Today, not when you finish the chapter.
Ask for an instance. Exercise 40.5. One meeting, four minutes, and it converts a criterion into a bar.
Score your own team. Exercise 40.8. Honestly, including the signals you do not have.
Take the unwanted job. Exercise 40.11 — the reconciliation, the metric definitions, the finance close. Case Study 1's engineer calls this the actual lesson of a decade.
Write the five-year letter. Exercise 40.14, with a calendar reminder.

A note on what to be skeptical of

Any career advice with a number of steps in the title. Including, in fairness, several of the framings in this chapter — which is why every weight in career_map.py is written down where you can argue with it.

Any advice that does not mention circumstance. Case Study 2's cycle 3 succeeded partly because the opportunities arrived. Advice that presents outcomes as purely a function of effort is selecting for people who got lucky and did not notice.

Any suggestion that a specific technology is "the future." §40.11: the products turn over every four to six years, reliably, and the confident predictions have a worse record than the technologies do.

And any framing — this chapter's included — that suggests there is one path. The tracks in §40.5 are five of many, the levels in §40.2 are one company's vocabulary, and the only claim here that is meant to be universal is the last one: that somebody has to be able to say whether the number is right, and that it may as well be you.