openkrill/hiring-signals-sample

Hiring Signals sample: one month of job postings from the top 1,000 companies

A free sample of the job posting data behind the Hiring Signals API. It holds the postings the 1,000 companies with the most postings put up between 2026-09-23 and 2026-10-22, read from their own public job boards and from open job feeds. Postings from sources that allow counts only are in the company counts, not in the postings file. For full titles, descriptions, every company we track and a daily refresh, use the API: https://tools.openkrill.app/tools/company-jobs

Uses

• Size up which companies hire most, in which job families, and how much of it is remote.
• Test a pipeline on real posting links and dates before you buy the full feed.
• Not for contacting people: the files hold no person's details.

Dataset Structure

Files

File	Rows	What it holds
postings.parquet, postings.csv	74,795	One row per posting.
companies.parquet, companies.csv	1,000	One row per company, with its posting counts and its public job board.

The Parquet and CSV files hold the same rows; manifest.json lists each file's row count, size and SHA-256. Built 2026-10-24 from a read of public job boards that finished 2026-10-23. Refresh: the sample is rebuilt from the export script, and the date above is its last build; the API refreshes daily.

Fields: postings

Field	Type	Meaning
posting_id	id	Our stable id for the posting: a hash of the source and the source's own posting id.
company	string	Company name as the employer's board or the feed gives it (the most common spelling).
company_rank	integer	The company's rank by postings in the window, 1 is the most.
source	string	Where the posting was read: the applicant tracking system or the job feed.
share	string	"full" when the source's terms allow redistribution, so its own fields are included; "links" when only the link and our labels may ship.
posted_date	datetime	The day the source says the posting went up (UTC, YYYY-MM-DD).
url	url	The posting on the employer's board or on the feed.
job_family	string	Our label from the title and department: software, data, design, product, marketing, sales, support, ops, finance, hr, assistant or other.
job_type	string	Our label: full_time, part_time, contract, internship or unspecified.
is_remote	boolean	Our label: true when the source or the location text marks the job remote.
title	string	Job title; only when share is "full".
department	string	Department; only when share is "full" and the source gives one.
location	string	Location text, places joined by "; "; only when share is "full".
country_codes	string	ISO country codes the source attaches, joined by ","; only when share is "full".

Fields: companies

Field	Type	Meaning
rank	integer	Rank by postings in the window, 1 is the most.
company	string	Company name.
postings	integer	Postings in the window from every source ruled keep, including sources that ship counts only.
postings_in_file	integer	Rows this company has in postings.parquet.
remote_postings	integer	Postings in the window our label marks remote.
top_job_family	string	The job family with the most postings.
sources	string	Sources the postings came from, joined by ",".
careers_url	url	The company's public job board on its applicant tracking system, when it has one.

Dataset Creation

Source data and what each source lets us share

Being allowed to read a source is not the same as being allowed to hand its posts to you. Each source below was checked for redistribution, and ships only what its terms allow: "full fields" sources grant redistribution, "link and our labels" sources are silent on it or limit it, so only the posting's link, date, company and our own labels ship, and "counts only" sources bar passing their posts on to other sites, so they appear only in the company counts. Sources marked "held" have site terms that name robots, scraping or copying; they are not read for this sample.

Source	What ships	Postings of the top companies	Rows in file	Ruling
Greenhouse	link and our labels	29,348	29,348	terms: "Job Board data is publicly available" (read right only; no redistribution grant).
Lever	link and our labels	14,368	14,368	terms: "These jobs may be scraped by third parties." (read right only; no redistribution grant).
Ashby	link and our labels	11,735	11,735	terms: "If you host your own careers page, you can use this data to populate it."
Workable	link and our labels	10,477	10,477	terms: Public job API for the employer's own careers page; no grant to redistribute the posts.
Recruitee	link and our labels	1,266	1,266	terms: Public job API for the employer's own careers page; no grant to redistribute the posts.
Personio	link and our labels	69	69	terms: Public job API for the employer's own careers page; no grant to redistribute the posts.
Teamtailor	counts only	1,675	0	terms: Terms bar users from "republish ... or distribute ... any material or information that you receive" from the Service.
Pinpoint	link and our labels	0	0	terms: "You can use our jobs feeds to help display your jobs on another website." (the employer's jobs, not ours to pass on).
Breezy HR	link and our labels	5,829	5,829	terms: Public job API for the employer's own careers page; no grant to redistribute the posts.
Rippling	link and our labels	0	0	terms: Public job API for the employer's own careers page; no grant to redistribute the posts.
Gem	link and our labels	143	143	terms: Public job API for the employer's own careers page; no grant to redistribute the posts.
ADP Workforce Now	link and our labels	0	0	terms: Legal page answers 403 to a plain read; no redistribution grant found.
Gupy	link and our labels	0	0	terms: Public job API for the employer's own careers page; no grant to redistribute the posts.
Keka	link and our labels	48	48	terms: Public job API for the employer's own careers page; no grant to redistribute the posts.
Paylocity	link and our labels	148	148	terms: Public job API for the employer's own careers page; no grant to redistribute the posts.
softgarden	link and our labels	1,352	1,352	terms: Public job API for the employer's own careers page; no grant to redistribute the posts.
Workday	not read (held)	0	0	terms: Site terms: "Use any data mining, robots or similar data gathering or extraction methods designed to scrape or extract data from our Sites" is barred.
BambooHR	not read (held)	0	0	terms: Terms bar use "with any robot, spider, other automated device, or manual process to monitor or copy any content".
SAP SuccessFactors	not read (held)	0	0	terms: "Automated extraction, copying, or indexing is not allowed without explicit permission".
Cornerstone OnDemand	not read (held)	0	0	terms: "You must not reproduce, distribute, modify ... republish, download, store or transmit any of the material on our Site."
Oracle Recruiting Cloud	not read (held)	0	0	terms: "the Materials may be used solely for your personal, informational, noncommercial purposes ... may not be redistributed."
Eightfold	not read (held)	0	0	terms: Terms bar "any robot, spider, or other automatic device, process, or means to access the Website".
Phenom	not read (held)	0	0	terms: Each board sits on the employer's own site under its own terms, which were not read one by one.
Remote OK	link and our labels	2	2	terms: "Please link back (with follow, and without nofollow!) to the URL on Remote OK and mention Remote OK as a source"; no redistribution grant.
Himalayas	counts only	240	0	terms: "Please do not submit Himalayas jobs to third-party websites".
Jobicy	full fields	10	10	terms: "You can store, cache, process, and redistribute job data as needed." Credit Jobicy and link each listing.
Arbeitnow	link and our labels	0	0	terms: "I would appreciate linking back to the site."; no redistribution grant.
Python.org Jobs	full fields	0	0	terms: Posters grant "all other users of the web site an irrevocable, worldwide, royalty-free, nonexclusive license to reproduce, distribute ... such content".
Startup Jobs	link and our labels	0	0	terms: "Reposting is welcome, as long as each listing has a dofollow link and a canonical tag pointing to the original"; a file cannot carry a canonical tag, so only the link ships.

Curation

1. Every public job board in our seed list (410,005 postings read) and each open feed was read at most one board per second per host, after its robots.txt.
2. Postings with no posting date (20,274) were left out, since they cannot be placed in the month.
3. A posting listed by an employer's board and by a feed counts once, from the board.
4. The 116,920 postings in the window came from 8,909 companies; the 1,000 with the most postings were kept.
5. Job family, job type and remote are our own labels from the title, department and location text.

Personal and Sensitive Information

No person's details are included. No recruiter, hiring manager or contact field is read from any source, the description text is not shipped, and a title, department or location holding an email address or a phone number is dropped. To have a row removed, open a discussion on the dataset page.

Licensing Information

Our compilation, the counts and our labels are under CC BY 4.0: credit "Hiring Signals by OpenKrill" with a link to https://tools.openkrill.app/tools/company-jobs. Rows whose share is "full" also carry the source's own terms: credit Jobicy or Python.org Jobs and keep the link to each posting. Rows from the feeds Remote OK, Arbeitnow and Startup Jobs link to the feed's own listing: name the feed as the source and keep that link when you show them. Company names and trademarks belong to their owners; the postings themselves stay on the employers' boards and the feeds, which the url field links.

Bias, Risks, and Limitations

• Only companies on the boards and feeds we read; large employers on held systems (Workday, SAP SuccessFactors, Oracle, BambooHR, Cornerstone, Eightfold, Phenom) are not in this sample.
• The seed list leans toward companies that hire remotely, since it was first built for our remote job index.
• A company with boards on two systems under two spellings may count as two companies.
• Some top boards belong to recruiting agencies or job aggregators that post on their own applicant tracking system; they count as the company that runs the board.
• Boards that list no posting date (Rippling, Pinpoint, most Gupy boards and Python.org Jobs) cannot be placed in the month, so they are left out.

Get the full data

The Hiring Signals API answers with every open posting of a company, its full text, and new and closed postings as they happen. This sample on Hugging Face: openkrill/hiring-signals-sample.

Citation

@misc{hiring_signals_sample_2026_10_22,
  title = {Hiring Signals sample: one month of job postings from the top 1000 companies (2026-09-23 to 2026-10-22)},
  author = {OpenKrill},
  year = {2026},
  url = {https://tools.openkrill.app/tools/company-jobs}
}