Intern Watch

Checks ~170 companies' job boards daily and emails you only when a new 2027 SWE intern posting shows up. Runs for free on GitHub Actions - no server, no computer that has to stay on.

Most companies are checked through their ATS's public JSON API (Greenhouse, Lever, Ashby, Workday, SmartRecruiters, plus Amazon's own search JSON and Eightfold boards like Netflix's - real titles, direct application links, ~1s each). Custom career sites fall back to a headless real-Chrome browser, where hits come from actual job links on the page ("link mode") rather than fuzzy text windows - window matching used to turn pagination controls and filter sidebars into "postings" and let a tech posting's keywords justify a non-tech neighbor.

Note: your doc had 21 companies, not 25 - Amazon, Microsoft, Meta, Apple, Google, Nvidia, Databricks, Stripe, Palantir, Uber, Airbnb, Salesforce, Adobe, Netflix, Oracle, LinkedIn, Roblox, Coinbase, Spotify, Snap, TikTok. If there are 4 more you meant to include, just add them to companies.json the same way (see Step 2).

How this version is different from a plain "fetch the page" script

I checked a few of your companies' pages directly first. None of them run on Greenhouse or Lever (the two ATS platforms with a clean public API) - they're all custom-built career sites, and several (Palantir in particular) only load their actual job listings via JavaScript after the page opens in a browser. A plain HTTP fetch sees an empty shell on those.

So monitor.py now uses a real headless browser (Playwright + Chromium) to load each page the way a person actually would, then searches the rendered text. It's still just a script - no LLM calls, no per-run cost beyond GitHub's free compute minutes.

Step 1: Create the GitHub repo

1. Go to github.com, click New repository. Name it something like intern-watch. Keep it Private. Skip adding a README (you already have one here).
2. Upload this folder's files via Add file → Upload files on the repo page (drag-and-drop works, including the .github folder), or push via git if you're comfortable with that.

Step 2: companies.json is already filled in

All 21 companies from your doc are already in there, pointed at the URLs you gave me, set to "ats": "browser" (the headless-browser mode). You don't have to do anything here unless you want to add more companies or swap in a more specific URL - see "If a company never finds anything" below.

Step 3: Set up email notifications

1. If using Gmail: turn on 2-Step Verification on your Google account, then create an App Password at myaccount.google.com/apppasswords. Use that (not your normal password) below.
2. In your GitHub repo: Settings → Secrets and variables → Actions → New repository secret. Add these:
   • SMTP_HOST → smtp.gmail.com
   • SMTP_PORT → 587
   • SMTP_USER → your Gmail address
   • SMTP_PASS → the app password from step 1
   • NOTIFY_EMAIL → where you want alerts sent (can be the same address)

Prefer Slack? Say so and I'll swap the email step for a one-line Slack webhook call.

Step 4: Test it before trusting it

1. Go to the Actions tab → Check Intern Postings → Run workflow (manual trigger).
2. Click into the run and watch the log - it prints which company it's checking as it goes, and any errors per-company.
3. The first run treats every existing 2027 posting as "new" since there's nothing to compare against yet - expect one slightly noisy email the first time, then only genuinely new postings after that.

Important: how to tell "nothing posted yet" from "broken check"

It's July 2026 - most of these companies haven't opened Summer 2027 intern applications yet, so a first run turning up zero hits across the board is expected, not necessarily a bug. To sanity-check that the rendering itself is actually working (as opposed to just finding nothing because nothing's posted):

1. Temporarily edit YEAR_PATTERN in monitor.py to match 2026 instead of 2027, and run the workflow manually.
2. If that turns up real 2026 intern postings for at least some companies, the pipeline itself is working - change it back to 2027 and let it run on schedule.
3. If it turns up nothing even for 2026, that company's page needs a different URL or extra handling (see below).

If a company never finds anything

A few of your URLs are program-overview pages rather than the actual job search/listing page (for example, a "why intern with us" page versus the page with actual role titles). Those will often render fine but just won't contain any job listings to match against. If a specific company keeps coming up empty even after the 2026 sanity check above, look for a "See open roles" / "View all jobs" link on that page and swap in that URL instead - or send it to me and I'll find the right one and adjust the entry.

A couple of others may load results only after you type into a search box or click a filter, which a plain page load won't trigger. If that turns out to be the case for a given company, flag it and I'll add site-specific handling (Playwright can click and type, it just needs a few extra lines per site).

The Extern calendar (summer 2027 release dates)

A second, separate watcher: extern_watch.py scrapes extern.com's per-company internship guides (the "Quick Facts" table each one carries) for every company with a tech program - Extern's own tech category plus tech tracks at non-tech companies (banks, retail, hospitality: think Wells Fargo Technology Summer Analyst) - and maintains:

• InternWatch.xlsx - the human-readable watchlist. Sheet 1 is sorted so whatever Extern expects to open soonest sits at the top (green = window open now, yellow = opening within ~2 months), with each company's career page link, Extern guide link, rolling policy, and pay. Sheet 2 holds companies with no formal intern program. Regenerated every run - don't edit it by hand.
• extern_calendar.json - the machine-readable state the diff runs against.

It runs daily (.github/workflows/extern-calendar.yml, ~9 AM ET). You only get an email when something actually changed: a window moved, a career link changed, a company gained/lost a formal program, or Extern published brand-new guides (those are listed for review, never auto-added - some are non-tech). Add or remove companies in extern_companies.json.

The Simplify live watch (fastest signal)

simplify_watch.py polls the community-maintained SimplifyJobs/Summer2027-Internships listings hourly (.github/workflows/simplify-watch.yml) and emails the moment a new posting appears - the community typically reports postings within hours of them going live. One HTTP request per run, stdlib only.

Filters (edit TECH_CATEGORIES etc. in the script): Summer 2027, Software + AI/ML/Data categories, US locations, and postings requiring U.S. citizenship are dropped (CPT). "No sponsorship" postings are kept with a note - CPT internships don't need sponsorship; conversion later would.

The underclassman watch

underclass_watch.py (runs with the Simplify watch, every 30 min) watches freshman/sophomore-specific sources - the underclassmen-opportunities and summer-2027-internships GitHub repos plus Extern's verified underclassmen program directory (STEP, Explore, Ignite, Propel, TEIP, ...) - and alerts on new rows. The full dataset renders as the "Underclassman programs" sheet in InternWatch.xlsx. Apply to these with the Class of 2029 graduation date.

How the layers fit together

Layer	Source	Lag	Job
Calendar	extern.com guides	days-weeks	when to expect each window, tracks, visa stance
Live watch	Simplify community list	hours	"a posting just went up" across hundreds of companies
Direct checks	company career pages	your cron interval	authoritative check on the shortlist in companies.json

Schedules run in two tiers. The BASE tier fits a private repo's free 2,000 Actions minutes/month: daily full sweep + fast lane every 2h + Simplify every 2h + daily Extern (~1,850 total). The MAX tier - fast lane every 20 minutes 24/7, Simplify every 30 minutes, three full sweeps a day - is scheduled too but its jobs only run when the repository is PUBLIC (unlimited free minutes); on a private repo GitHub skips them at zero cost. Making the repo public upgrades the cadence automatically, no config change needed.

Adjusting things later

• Schedule: edit the cron: line in .github/workflows/check-intern-postings.yml. crontab.guru helps build the expression.
• Keyword matching: edit KEYWORD_PATTERN in monitor.py if a company phrases the role unusually.
• Add/remove companies: edit companies.json and push.

Costs

Free. Loading 21 pages in a headless browser plus a bit of setup time comes to a few minutes per run - even running daily for a full month stays well within GitHub Actions' free 2,000 minutes/month.