Careers in Data Science

The jobs that make up data science and what each involves, as an orientation to the collection. Where the title "data scientist" came from; data engineers, analysts, data scientists, machine learning engineers and statisticians placed along the stages of a project, with the articles that teach each role's core skills; official U.S. Bureau of Labor Statistics figures for data scientists and related occupations (employment in 2025, median pay in May 2025, projections for 2025 to 2035), what such projections mean and why they change; and the skills that the sources agree on.

When DJ Patil and Jeff Hammerbacher began, in 2008, to compare notes on the data teams they were building at LinkedIn and Facebook, they needed a name for the people they hired. "Business analyst" seemed too limiting. "Data analyst" was a contender, but "we felt that title might limit what people could do". "Research scientist" suggested work done in labs isolated from the products. "The term that seemed to fit best was data scientist: those who use both data and science to create something new" (Patil, 2011).

The title was chosen to be broad, and it has stayed broad. "Data scientist" names a specific job, and it is also used loosely for everyone who works with data, from the engineers who build the pipelines to the statisticians who design the studies. This article separates the main roles, places each along the stages of a project described in The Data Science Lifecycle, points to the articles in this collection that teach the core of each, and reports what official labor statistics say about the occupations, with their dates. Titles and their boundaries vary from one employer to the next; the descriptions below are of kinds of work, not of any organization's job ladder.

The roles

Data engineers

A data engineer builds and runs the systems that collect data, store it, move it between systems and turn it into tables that others can use. The work includes pipelines that extract data from operational systems, transform it and load it into an analytical store, and the warehouses and lakes that hold it, all of which Data Engineering introduces from an analyst's point of view. Data engineering sits closest to software engineering and databases: Relational Databases, the SQL articles from Simple SQL Queries to Queries in Python, and file formats and schemas are its daily material.

The work matters to everyone downstream because the data an analyst receives are shaped by decisions made upstream: which columns exist, what types they have, how fresh they are, and which values were silently converted along the way. The zero that a file reader turns into $5.4 \times 10^{-79}$ in Cleaning Data is a small example.

Data analysts

A data analyst summarizes, visualizes and interprets data to answer specific questions, usually for people who will make a decision with the answer. The emphasis is on description: what happened, how groups differ, how a quantity has changed. The core skills are those of the early and middle topics of this collection: SQL for retrieval, groups and pivot tables for summaries, honest charts, and enough statistics to know when a difference is real (Descriptive Statistics, Hypothesis Testing). Analysts who work mainly with financial and business performance data are often called business intelligence analysts, and work in dashboard and reporting tools as well as in code, which Tools for Visualizing Data surveys.

Data scientists

The data scientist, in the narrower sense of the title, covers the middle of a project: framing a question so that data can answer it, assembling and exploring the data, choosing and fitting models, judging them on held-out data, and explaining what they show. The Bureau of Labor Statistics (BLS) lists the duties of data scientists as determining "which data are available and useful for the project"; collecting, categorizing and analyzing data; creating, validating, testing and updating "algorithms and models"; using data visualization software to present findings; and making "business recommendations to stakeholders based on data analysis" (BLS, 2026a).

In Donoho's (2017) terms, a data scientist uses "tools and viewpoints from both of Leo Breiman's modeling cultures": generative models, such as the regressions of Introduction to Regression, whose parameters say something about how the data arose, and predictive models, such as random forests and neural networks, judged only by how well they predict. Most of this collection, from Performing Exploratory Data Analysis through Model Selection, is the data scientist's core.

Machine learning engineers

A machine learning engineer turns a model into software that runs reliably in production: serving predictions at the required speed and volume, retraining as data change, testing, and monitoring. The size of that work is easy to underestimate. Sculley et al. (2015), engineers at Google, observed that "only a tiny fraction of the code in many ML systems is actually devoted to learning or prediction"; the infrastructure around it is, in the caption of their figure, "vast and complex". They also found that while "developing and deploying ML systems is relatively fast and cheap", maintaining them over time "is difficult and expensive". The deployment and monitoring phases of The Data Science Lifecycle are this role's territory, along with the software practices of Data Science Packages and Python Functions.

Statisticians

Statisticians design studies and analyses, and derive and check the methods that the other roles apply. The BLS groups them with mathematicians, who "analyze data and apply computational techniques to solve problems" (BLS, 2026b). Their work is the subject of the Applied Statistics course pages on this site, from Descriptive Statistics to Analysis of Variance, and the overlap with data science is large. Donoho (2017), reviewing presentations on data science, could not avoid noticing "that the underlying tools, examples, and ideas which are being taught as data science were all literally invented by someone trained in Ph.D. statistics, and in many cases the actual software being used was developed by someone with an MA or Ph.D. in statistics". How the two fields came to overlap is traced in A Historical Overview of Data Science.

How the roles divide a project

Phase of a project Data engineer Data analyst Data scientist ML engineer Statistician
Framing the question ● ● ●
Collecting and storing data ● ● (study design)
Preparing and exploring data ● ● ● ●
Modeling ● ● ●
Evaluating ● ● ● ●
Deploying and monitoring ● ●
Communicating results ● ● ●

The table is a sketch of where each role's center of gravity lies, not a rule. In a small organization one person covers most of the rows; in a large one, several people share each.

What the labor statistics say

The most authoritative public source on U.S. occupations is the BLS Occupational Outlook Handbook, which publishes for each occupation its employment, its median pay and a ten-year projection. It has a profile for data scientists. Data engineering and machine learning engineering do not have profiles of their own; the nearest are database administrators and architects, whose work is to "create or organize systems to store and secure data", and computer and information research scientists, whose profile mentions experiments "frequently using techniques from data science and machine learning" (BLS, 2026c, 2026d). The figures below were read from the Handbook in November 2026; the pages were last modified in August and October 2026.

Occupation (BLS) Jobs, 2025 Median annual pay, May 2025 Projected change, 2025 to 2035 Typical entry-level education
Data scientists 275,600 120,230 dollars +35% Bachelor's degree
Statisticians 31,300 105,650 dollars +11% Master's degree
Operations research analysts 113,100 88,940 dollars +12% Bachelor's degree
Database administrators and architects 144,500 126,760 dollars +4% Bachelor's degree
Computer and information research scientists 38,600 140,300 dollars +22% Master's degree

Sources: BLS (2026a, 2026b, 2026c, 2026d, 2026e).

The education entry for statisticians is the one the BLS gives for mathematicians and statisticians together. For comparison, the median annual wage for all workers in May 2025 was 50,980 dollars (BLS, 2026a). The BLS describes the projected growth for data scientists as "much faster than the average for all occupations" and expects about 24,800 openings a year, on average, over the decade, many of them "expected to result from the need to replace workers who transfer to different occupations or exit the labor force" (BLS, 2026a). The largest employers of data scientists in 2025, by industry, were publishing, broadcasting and content providers (12%), insurance carriers and related activities (10%), computer systems design (10%), and management of companies and enterprises (10%).

Reading these numbers

Three cautions apply to figures like these.

  • A median is a midpoint, not a typical offer. "The median wage is the wage at which half the workers in an occupation earned more than that amount and half earned less" (BLS, 2026a). It mixes beginners with senior staff and every region of the country.
  • A projection is a model's output for a stated period. These figures are for 2025 to 2035, and each page carries the date it was last modified. A growth figure quoted without its period, or copied from an earlier edition, cannot be compared with them.
  • The categories are not the job titles in advertisements. An employer's "data engineer" may be counted under database architects, software developers or data scientists, depending on how the job is classified. Employment counts for the narrower titles cannot be read off these tables.

The figures describe the United States. Other countries publish comparable statistics under their own occupational classifications.

Skills the sources agree on

Lists of data science skills are easy to find and hard to verify. Three sources used elsewhere in this collection agree on a short core.

  • Computing with data. Donoho's (2017) "greater data science" includes computing as one of six divisions, and expects a data scientist to "know and use several languages for data analysis and data processing". In this collection that means Python with NumPy and pandas, and SQL.
  • Exploration, preparation and visualization. Donoho notes that "some say that 80% of the effort devoted to data science is expended" on getting to know messy data and making it ready, and puts exploration and presentation beside modeling. The Data Wrangling and Data Exploration topics cover this work.
  • Modeling with a sense of its limits. Both of Breiman's cultures, with the evaluation habits of Training, Validation, and Test Sets and Cross-Validation.
  • Knowledge of the domain. Hernán, Hsu and Healy (2019) argue that causal questions, the ones decision makers usually care about, require "domain expert knowledge" as well as data and algorithms, and Why Data Science shows a simulated case where a good predictor gives the wrong answer to such a question.
  • Communication. Patil's (2011) list of the traits he looked for when hiring adds "curiosity" and "storytelling", "the ability to use data to tell a story and to be able to communicate it effectively", to technical expertise.

Every role uses all of these to some degree. What separates them is the proportion: an engineer spends more time on the first, an analyst on the second and the last, a data scientist on the third, and a statistician on the third and fourth together.

The same work also carries obligations to the people the data describe, which no list of technical skills includes. Ethics in Data Science covers them.

References