A job post is one of the few public documents where a company says which tools it runs. "3+ years administering Okta and Intune" tells you more about an IT stack than most press releases. I wanted a quick way to ask "which of these companies is hiring for Okta or Intune right now?", so I wrote a small tagger and ran it against a handful of public job boards.
Every number below comes from one sample I ran on 2026-11-02 between 9:36 and 9:38 a.m. ET: 22 public boards, 4,330 open postings, 28 HTTP requests. It is a small, hand-picked sample of mostly tech companies, so read the numbers as "what this sample showed", not as a picture of the market.
Where the posts come from
Greenhouse, Lever and Ashby each publish a JSON feed of an employer's open jobs, the same data their hosted careers pages use. No login, no scraping of HTML pages:
Greenhouse boards-api.greenhouse.io/v1/boards/<slug>/jobs?content=true
Lever api.lever.co/v0/postings/<slug>?mode=json&limit=100&skip=0
Ashby api.ashbyhq.com/posting-api/job-board/<slug>
The slug is the one in the company's careers URL. Greenhouse returns every job with its description in one request; Lever pages 100 at a time; Ashby returns the whole board.
I kept the fetcher slow: it reads robots.txt on each host first, waits at least a second between requests and sends a User-Agent that names the project. The 2026-11-02 sample was 15 Greenhouse boards (2,699 postings), 2 Lever boards (387) and 5 Ashby boards (1,244).
Tagging, and the false positives
The tagger is a hand-written dictionary of 328 tools (Okta, Microsoft Intune, Jamf, CrowdStrike, Kandji and so on), each with its spellings. "Intune", "Microsoft Endpoint Manager" and "MEM" all map to Microsoft Intune. The first version was a substring search. Here are three things the 2026-11-02 sample showed about why that does not work.
1. Substrings. In the 2026-11-02 sample, 2,154 of the 4,330 postings contain the letters "aws". Only 279 are tagged AWS; "laws" alone shows up in 1,983 postings. "rust" is worse: 2,695 postings contain it, 201 are tagged Rust, and nearly all of the rest are "trust", "trusted" and "entrusted". The fix is whole-word matching: a spelling only counts if no letter, digit or underscore touches it on either side.
import re
B, E = r'(?<![A-Za-z0-9_])', r'(?![A-Za-z0-9_])'
re.search(B + r'aws' + E, 'state and local laws', re.I) # None
re.search(B + r'aws' + E, 'AWS, GCP or Azure', re.I) # match
Everyday words get one more rule: Rust, Slack, Zoom, Git and Notion only match with their usual capitals. I dropped bare "Go" entirely. In this sample 3,755 postings contain "go" as a substring ("goals", "government", "good"), and as a whole word it is also an ordinary verb. Only "Golang" and "Go (Golang)" count, which means "Python or Go" is missed. I took that trade.
2. The employer's own name. All 811 OpenAI postings in the sample mention OpenAI, and all 400 Datadog postings mention Datadog. Counting those says nothing about the stack, so the tagger drops the employer's own product. In the 2026-11-02 sample that rule removed a tag from 2,273 postings across 8 companies.
3. Culture lines. Slack shows up in 78 postings in the sample, often in lines like "coffee chats and Slack conversations that help them stay connected" or "works mostly async, with regular pairing sessions, 1:1s, and Slack". That is how a team talks, not a skill being hired for. Slack, Zoom, Microsoft Teams and Webex are now gated: they only count inside a requirements section or near a phrase like "experience with". After gating, 38 of the 78 Slack postings still count. This rule is a heuristic, since it depends on headings, and it will get some posts wrong.
There is also a rule that skips "trusted by companies like..." and "backed by investors..." sentences. It exists because those sentences name tools, but it did not fire on any of the six identity and device tools I counted below in the 2026-11-02 sample.
From posts to a table
With tags per posting, the per-company table is a count of open postings that mention each tool:
jobstack boards greenhouse:discord greenhouse:asana ashby:openai ... \
--focus "Okta,Microsoft Intune,Jamf,CrowdStrike,Kandji,Microsoft Entra ID" \
--jobs-out jobs.jsonl
Part of the table from the 2026-11-02 sample: the nine boards with the most hits for these six tools. Seven more boards had one or two Okta postings each (Notion and Vanta 2; Coinbase, Duolingo, Figma, Reddit and Robinhood 1).
| board | open postings | Okta | Intune | Jamf | CrowdStrike | Kandji | Entra ID |
|---|---|---|---|---|---|---|---|
| Cloudflare (Greenhouse) | 467 | 4 | 1 | 2 | 0 | 1 | 4 |
| Palantir (Lever) | 301 | 6 | 0 | 0 | 0 | 0 | 6 |
| Asana (Greenhouse) | 110 | 4 | 1 | 1 | 3 | 0 | 0 |
| OpenAI (Ashby) | 811 | 0 | 3 | 3 | 0 | 0 | 2 |
| Ramp (Ashby) | 173 | 2 | 0 | 2 | 1 | 1 | 0 |
| Discord (Greenhouse) | 41 | 2 | 2 | 2 | 0 | 0 | 0 |
| Dropbox (Greenhouse) | 24 | 1 | 2 | 0 | 0 | 0 | 3 |
| Airbnb (Greenhouse) | 201 | 3 | 0 | 0 | 0 | 0 | 0 |
| GitLab (Greenhouse) | 235 | 2 | 0 | 1 | 0 | 0 | 0 |
Some takeaways from the 2026-11-02 sample:
- Okta appeared in 33 postings at 15 of the 22 companies. Some of those are the same job listed in several locations; counting each company and title once, it is 26.
- Microsoft Intune appeared in 9 postings at 5 companies. Only 3 postings mentioned both Okta and Intune: one at Cloudflare and two at Discord.
- The hits sit in a few kinds of roles: IT support and IT ops, corporate security, identity engineering, and customer-facing support for security products. A Discord IT Ops post lists "identity (Okta), device management across both Mac (Jamf) and Windows (Intune)"; an Asana IT Support post asks to "build and deploy new laptops using standard deployment tools like JAMF, Intune".
- 6 of the 22 boards (Datadog, Instacart, Linear, Lyft, Pinterest and Spotify) had no posting that mentioned any of the six tools.
A count of open postings is a weak signal on its own: a company with 800 openings will mention more tools than one with 40. It is most useful as a list of specific postings to read, which is why --jobs-out saves every tagged posting with its URL.
The code
The code is in my repo, MIT-licensed: the tagger, the dictionary, the three board readers and the tests. https://github.com/fecklesspleb/jobstack
from jobstack import Tagger
Tagger.default().tag("IT Support Specialist",
"Build and deploy new laptops using tools like JAMF, Intune, ABM")["tags"]
# ['Jamf', 'Microsoft Intune']
If you try it on other boards and find a tool it misses or a false positive it lets through, an issue with the posting text is the most useful thing you can send.
I wrote this post and the code with AI assistance; the numbers come from the sample run described above.
Top comments (0)