PeopleAdmin Jobs API

PeopleAdmin is the applicant tracking system US higher education runs its hiring on. Not tech companies, not startups: universities and colleges, with the posting mix that implies — faculty appointments, lab technicians, groundskeepers, associate deans, adjuncts. If you're building anything that touches academic hiring, this is the platform your data lives on, and it's a pool almost nobody aggregating startup jobs ever looks at.

Every institution gets a tenant at <tenant>.peopleadmin.com, and every tenant publishes an Atom feed with the full posting body inline. One unauthenticated GET, no auth, no detail pass. On the university board we tested that was 60 postings, all 60 with descriptions.

How PeopleAdmin boards actually work#

One GET, no auth, the whole board:

curl "https://<tenant>.peopleadmin.com/postings/search.atom"

<tenant> is the subdomain — an institution's own short name. Each <entry> carries <title>, <content> with the complete posting body, <updated> as an ISO 8601 timestamp with a zone, and a link to /postings/<id>. The ids are tenant-local, so they need the tenant folded into any dedup key or two institutions' posting 5453 will collide.

That's the whole surface. It's a well-behaved feed, and it has three traps.

What bites you#

.json is a trap. The same path with a .json extension looks like it should work and instead 302s to /404. Only .atom is live. Follow the redirect without checking where you landed and you're parsing an error page.

?page=N is silently ignored, and that looks like a cap. Pass a page parameter and you get back a byte-identical document every time. The natural reading is that the feed caps out around 150 entries and the paging is broken — so you go hunting for the parameter that unlocks the rest, and there isn't one.

It isn't a cap. We checked it the only way worth checking: walking the board's own HTML pagination to exhaustion and comparing the id sets.

The feed returns the complete set in one document, which is why it has no paging controls. It doesn't need them. (Not proven above ~150 — no tenant we've tested is bigger — so a very large institution reporting exactly N entries across many HTML pages is worth re-checking rather than trusting.)

There is no location. Anywhere. Not in the feed, not on the detail page. We checked both. Higher-ed postings state where the job is inside the body prose, in whatever form the department felt like typing, and the structured field simply does not exist on this platform. Zero percent is the honest number, and emitting a guess would be worse than emitting nothing — a wrong location is unfalsifiable downstream, where a missing one is at least visible.

The obvious repair is the wrong one, too. You cannot default the country to the US just because PeopleAdmin is a US higher-ed platform: it has Canadian tenants. Vancouver Community College and Capilano University are both real boards here, and stamping United States across them would be confidently, silently wrong on every posting. Location on this source has to come from the body or from the institution's own company profile, not from an assumption about the vendor's market.

Or use ours#

The feed itself is easy — one GET, no auth, the complete set in a single document — so for a few institutions you should just read it, minding the .json redirect and not mistaking the ignored ?page=N for a cap.

?source=peopleadmin is served, and what it adds is the pool rather than the parse. PeopleAdmin publishes no index of its tenants, and academic hiring is not a board list anyone aggregating startup jobs has assembled — faculty appointments, lab technicians and adjuncts sit in the same Job shape as a Greenhouse engineering role, on /ats, queryable with the same filters.

Location is the honest limit and normalisation can't rescue it. The platform has no location field anywhere, so what's available is the body prose and the institution's own company profile — not a guessed country, and specifically not the United States default that the Canadian tenants would make silently wrong on every posting. Filter this source accordingly.

/v1/meta → sources remains the authority on what you can filter on: source is a csv filter, not an enum, so a miscased ?source=PeopleAdmin returns an empty array rather than a 400 telling you what went wrong.