2027 AI College Jobs: four Markdown files refreshed daily

Four Markdown files carry the entire data set

The repository is a data set, not a program. The top level holds four Markdown files and one workflow directory: README.md for the USA internship list, NEW_GRAD_USA.md, INTERN_INTL.md and NEW_GRAD_INTL.md, plus .github/. There is no source tree, no package manifest, no build script and no licence file.

The README states the scope in one sentence: a list of AI/ML and Data Science jobs for college students in search of internships or new graduate positions, updated daily, with jobs posted within the last 120 days prioritized.

That claim is checkable from outside the repository rather than from the files themselves. The last push to the default branch is dated 2 November 2026, and there are no releases at all, so there is no version history to compare against. No primary language is detected and the licence field is empty, which is consistent with a repository that holds nothing but Markdown and HTML anchors.

654, 501, 613, 501 and the same three buckets everywhere

The four headline counts are the fastest way to understand the shape of the data. USA internships: 861 available. USA new graduate: 548 available. International internships: 781 available. International new graduate: 711 available.

Each of those four lines splits into the same three buckets, labelled FAANG+, Quant and Other, so one taxonomy applies whether a role sits in Menlo Park or somewhere else entirely. Each bucket link is an anchor into the file the headline points at.

The navigation is slightly asymmetric, and it matters when you are trying to browse. The internships link in the USA block points at the repository root, so the README doubles as its own internships index and the FAANG+ table appears inline. The other three headline links point at separate files, which means new graduate roles and every international listing are one click away while internships are read in place.

Six columns per row, and an age counter that ages in place

Every row carries the same six columns: Company, Position, Location, Salary, Posting and Age. The first visible row is Meta, Research Scientist Intern - Video RecSys - PhD, Menlo Park, CA, $50/hr, three days old. The next is Meta, Research Scientist Intern - Computer Vision - Video Intelligence - PhD, Bellevue, WA +1, $50/hr, also three days old.

The Age column is what makes the 120 day rule visible. Values in the visible rows run from 3d through 4d, 5d and 6d, all well inside the window.

What the columns do not include matters just as much. There is no closed or filled marker, no application deadline column and no status field, so a row records a posting rather than an open role, and nothing in the table tells you whether anyone applied.

Pay is one hourly rate per row, and the visible range is $50 to $62

Money is expressed as an hourly rate rather than an annual figure, which matters when you compare these rows against the new graduate files. In the visible FAANG+ entries the rates run from $50/hr at Meta, through $52/hr at Microsoft, $54/hr at Salesforce and $60/hr at Stripe, up to $62/hr for an NVIDIA internship.

The Salary column carries a single number for the whole row, and the README sets out no rule for handling a posting that advertises a range, so it is not possible to tell from the repository alone whether ranges are collapsed to one figure or skipped.

Location has the same character. California, USA +3 appears for a Salesforce internship and Bellevue, WA +2 for a Meta role, and the +N notation for extra locations is left undefined.

Every posting link leaves for a different applicant tracking system

The Posting column is a column of outbound links to employer-owned application systems, and the destinations are not uniform. Microsoft rows point at apply.careers.microsoft.com, Meta at metacareers.com, NVIDIA at nvidia.wd5.myworkdayjobs.com, Salesforce at salesforce.wd12.myworkdayjobs.com, and Stripe at search URLs carrying a gh_jid parameter.

Two consequences follow. Pulling these rows into a comparable internal record means handling at least five different application flows rather than one field to copy. And a dead row stays invisible until someone opens the link, because Age is the only freshness signal in the table.

That is also why the daily update claim carries weight. The README describes the positions as updated daily, but nothing in the repository describes what happens to a row once a posting is pulled by the employer.

PhD fills more than half the visible FAANG+ rows

The position titles carry the degree requirement, which is the most useful filter the table hands you for free. In the visible portion of the FAANG+ block, PhD appears in more than half the rows: nearly all of the Meta Research Scientist Intern roles, Meta's ML Innovation for Products PhD Research Intern, and Stripe's PhD Machine Learning Engineer - Intern.

Master's shows up in one row by comparison, Stripe's Master's Machine Learning Engineer - Intern, while several titles name no degree at all, including Salesforce's AI Labs Summer Internship, Microsoft's Applied Scientist and Research Intern roles, and NVIDIA's VLSI CAD Timing Intern.

Nothing in the repository groups rows by degree or by seniority, so that filtering happens after the fact, in whatever spreadsheet you copy the table into.

The remaining space goes to promotion, not to data access

The rest of the README is promotion rather than plumbing. It links speedyapply.com, a Discord invite, a Chrome Web Store extension for job applications, and a second repository for software engineering roles at 2027-SWE-College-Jobs. The extension is the difference between a list you read and a list you act on, and it lives outside the repository on the web store.

What the repository does not offer is any structured form of the same data: no JSON export, no CSV, no documented feed, no API, and no stated rule for when a removed posting is deleted from a file. There is no recorded licence either, so the reuse terms for a list you did not assemble yourself are simply unstated.

Treat the Markdown as a snapshot. Copy it, check the links, and keep your own record of what you applied to, because none of that bookkeeping happens in the repository.

Editorial conclusion

The list suits a student who wants a current shortlist of 2027 AI/ML roles and is willing to work from a Markdown table rather than a job board with filters. It does not suit anyone who needs structured data, a searchable feed or stated reuse terms, because none of those exist here. Before you build anything on top of it, copy the tables into your own tracker, open each posting link to confirm the role is still live, and settle the reuse question with the maintainers rather than assuming a licence covers you.

Frequently asked questions

Official sources

1. Issues
2. Project website
3. README
4. speedyapply/2027-AI-College-Jobs on GitHub