Global Internship Finder

Discover fresh internships across Europe and beyond — collected automatically from public job boards and company career pages, then normalized, deduplicated, and relevance-scored.

Select countries, cities, roles, languages, and remote/hybrid/on-site preference, and get a clean, ranked dataset of validated internships — ready to power an internship search website.

Responsible use: Uses only public / authorized sources and official-style JSON endpoints (The Muse API, Arbeitnow API, Remotive API, and public Greenhouse/Lever ATS boards). It does not log in, scrape LinkedIn/Indeed, bypass CAPTCHAs, or access private/paywalled data. LinkedIn & Indeed are intentionally excluded because their terms/anti-bot controls prohibit it.

---

Sources (modular — add more by dropping a connector in src/sources/)

Source	Coverage	Notes
The Muse	Global, explicit Internship level	Best structured internship source (category + location)
Arbeitnow	Europe (esp. Germany)	Public EU job board
Remotive	Remote	Remote internships
Greenhouse	Per company	Public ATS boards — exposes deadlines & education
Lever	Per company	Public ATS boards

Input (all optional — run with none for a broad EU sweep)

Field	Description
countries / cities	Location filters, e.g. ["Germany","Netherlands"], ["Berlin"].
categories	Roles, e.g. ["Software Engineering","Data Science","Finance"] (also queries The Muse).
keywords	Must appear in title/description.
locationType	any / remote / hybrid / onsite.
languages	Required languages, e.g. ["English","German"].
sources	Which connectors to run. Default ["themuse","arbeitnow","remotive"].
greenhouseCompanies / leverCompanies	Company board tokens for ATS sources.
maxPerSource	Cap per source (default 150).
minRelevanceScore / maxAgeDays	Quality/freshness filters.
previousDatasetId	Prior run's dataset → new/expired/deadline-change detection.

Example inputs

{ "countries": ["Germany","Netherlands"], "categories": ["Software Engineering","Data Science"] }

{ "sources": ["greenhouse","lever"], "greenhouseCompanies": ["stripe","gitlab"], "leverCompanies": ["ramp"], "keywords": ["data"] }

{ "countries": ["France"], "locationType": "remote", "previousDatasetId": "<PREV_DATASET_ID>" }

Output

1) Internships (default dataset — active, ranked by relevance)

{
  "id": "3f9a1c22b7e04d18",
  "jobTitle": "Data Science Internship (m/f/d)",
  "company": "Example GmbH",
  "country": "Germany", "city": "Berlin",
  "locationType": "hybrid",
  "industry": "Data Science", "department": "Data",
  "internshipType": "internship", "duration": "6 months",
  "salary": null,
  "requiredSkills": ["python","sql","machine learning"],
  "educationRequirements": "Bachelor",
  "experienceRequirements": null,
  "languageRequirements": ["English","German"],
  "applicationDeadline": null,
  "postingDate": "2026-08-20T00:00:00Z",
  "sourceWebsite": "themuse.com", "source": "themuse",
  "sources": ["themuse","arbeitnow"],
  "sourceUrls": [{ "source":"themuse","url":"https://..." }],
  "url": "https://...", "companyUrl": null,
  "description": "…",
  "relevanceScore": 82,
  "relevanceReasons": ["country match","role match","keywords: python"],
  "classifications": ["new","high-match"],
  "status": "active",
  "collectedAt": "2026-08-30T18:00:00Z",
  "lastVerifiedAt": "2026-08-30T18:00:00Z"
}

2) Changes (named dataset changes)

{ "id":"3f9a1c22b7e04d18", "changeType":"new", "jobTitle":"…", "company":"…", "detectedAt":"…" }

changeType ∈ new · expired · deadline-changed · updated. Run summary is in key-value store OUTPUT.

Relevance score & classifications

Base 40 (already a validated internship) + country/city/role/keyword/language/location-type matches + freshness + multi-source corroboration → 0–100. Classifications: new (≤7d), deadline-soon (≤14d), remote, high-match (≥75), likely-expired (deadline passed or >90d old). Likely-expired listings are kept out of the active dataset but still surface in changes for historical tracking.

Scheduling & history

Schedule daily (or several times/day) and pass the previous run's dataset ID as previousDatasetId. Each run emits new/expired/deadline-changed/updated events, so you can maintain a live + historical view.

Website / database

db/schema.sql is a ready PostgreSQL schema with proper indexes (country, city, company, industry, location_type, deadline, posting_date, relevance, GIN on skills/languages/classifications) plus internship_sources, internship_changes, and saved_searches (alerts). Load the default dataset into internships (upsert on id, pattern included in the SQL) and the changes dataset into internship_changes. The website then queries/filter/paginates against these tables. The website + query API is a separate build from this Actor.

Reliability

Per-source fail-safe (one dead source never stops the run), retries with backoff, concurrency across sources, HTML→text normalization, and deterministic ids for stable dedup/upsert.

Limitations

• Internship detection is keyword/heuristic (multi-language: EN/DE/FR/ES/IT/NL/PT/PL) — very accurate on explicit sources, best-effort on free-text boards.
• Country/city are parsed from location strings via a curated hint list (extend COUNTRY_HINTS in normalize.js).
• Deadlines exist mainly on Greenhouse; other sources rarely publish them.
• LinkedIn/Indeed are out of scope by design (ToS/anti-bot).