AI researchers by country, subject and year

The files

A manifest at /data/index.json lists the files, their columns, their row counts and the date of the figures, so a script can check for a new version without downloading the tables.

How to download

In a browser, click CSV or JSON in the table above. The file saves to your downloads folder and opens in Excel, Numbers or any text editor.

From a terminal, with curl:

curl -O https://foundinghires.tech/data/countries.csv
curl -O https://foundinghires.tech/data/years.json

In Google Sheets, in any cell:

=IMPORTDATA("https://foundinghires.tech/data/countries.csv")

In Python, with pandas:

import pandas as pd
countries = pd.read_csv("https://foundinghires.tech/data/countries.csv")
years = pd.read_json("https://foundinghires.tech/data/years.json")["rows"].pipe(pd.DataFrame)

The CSV files are UTF-8 with a header row and a comma separator. The JSON files hold the same rows under a rows key, with the licence, the citation, the as-of date and the comparison windows alongside.

How to cite

The licence is Creative Commons Attribution 4.0. It asks one thing: say where the figure comes from. In a chart caption, a footnote, an article or a slide, write:

Source: Founding Hires, Atlas of AI research, https://foundinghires.tech/data.html

Add the as-of date printed in the file when the figure may be compared with a later one. Online, a link to this page is the citation. If you publish something built on these files, we are glad to hear about it from the contact address.

What the columns mean

• researchers: people with at least one indexed AI paper, counted once each. Country is the country of their institution on their most recent papers.
• early_career: researchers whose record reads as early career, typically a doctorate in progress or recently finished.
• in_industry: researchers whose institution is classified as a company by the source.
• newcomers_recent_window and newcomers_previous_window: researchers whose first indexed paper falls in each of two consecutive windows. The years of the windows are in the JSON files and in the manifest.
• subject: a topic from the OpenAlex classification of the literature. A researcher can carry several, so subject counts overlap.
• year and newcomers: researchers whose first indexed paper falls in that year, for the world, a country or a subject as the scope column says. The current year is incomplete.

Rows with fewer than twenty researchers are withheld. Definitions and limits are on the method page.

Common questions

Is the data free to use?

Yes, under a Creative Commons Attribution 4.0 licence. You can copy, reuse and republish the figures, including commercially, as long as you cite Founding Hires and link to this page.

How do I cite Founding Hires?

Write: Founding Hires, Atlas of AI research, foundinghires.tech/data.html, with the as-of date printed in the file. A link to this page in an article, a chart caption or a footnote is the citation.

Does the data contain names or contact details?

No. Every file is an aggregate: counts of people by country, subject or year. Rows with fewer than twenty researchers are withheld. The index itself, with names, is not a public dataset.

How often is it updated?

The index is refreshed nightly and the files are republished weekly, on the same schedule as the figures on the site. A file only changes when its numbers change, and each file carries the date of the figures it holds.

Why do the subject counts not add up to the country total?

A researcher can carry several subjects, so the counts by subject overlap. Country totals count people once.

Can I get a different cut of the data?

Write to us from the executive search page. Aggregates that keep every cell above twenty people can be added; anything that identifies a person cannot.