Job Postings AI Classifier – Seniority, Remote & Visa

AI job classification for LinkedIn, Indeed and any job scraper dataset: turn raw job postings into clean, filterable tags (seniority, job function, remote / hybrid / on-site, where remote work is allowed from, employment type, visa sponsorship, relocation support, salary stated, people management, degree requirement, posting language) and the minimum years of experience asked for.

It works on any job dataset — the output of our Greenhouse, Lever & Ashby Jobs Scraper or Remote Jobs Scraper, any other job scraper on Apify, or postings you paste in. Common field names (title, jobTitle, description, descriptionHtml, description_text, companyName, …) and nested ones (title.text, description.text) are detected automatically, so datasets from popular LinkedIn and Indeed job scrapers work without settings.

Every tag comes with a confidence score, so you can automate the clear cases and review the few uncertain ones.

Who uses it

• Job boards and aggregators — consistent seniority, remote and function filters across sources that all label things differently.
• Recruiters and sourcers — find the senior remote roles that sponsor visas without reading 2,000 postings.
• Labour-market research — count roles by function, level and work mode over time.
• Job-alert and newsletter builders — route postings to the right audience automatically.
• AI agents — a cheap, typed "what kind of job is this?" tool (Apify MCP).

How to use it

1. Run a job scraper (or pick any dataset with job postings), or paste postings into Or paste job postings.
2. Select the dataset(s) under Job datasets to tag and start the run.
3. Download the tagged postings as JSON, CSV or Excel, or join them back to your data by sourceIndex / id / url. Turn on Keep the source row's job fields (keepOriginalFields) to get the source row's job fields (posted date, salary, employment type, links, …) in original on every row, with no join needed. People's names and contact details are never copied. You can also chain it after a scraper with an Apify integration: use {"datasetId": "{{resource.defaultDatasetId}}"} as the input.

Input example

{
  "datasetIds": ["YOUR_DATASET_ID"],
  "maxItems": 5000
}

Or with postings inline (objects or plain text):

{
  "items": [
    {
      "title": "Senior Backend Engineer (Python)",
      "companyName": "Northwind Analytics",
      "location": "Remote (EU time zones)",
      "description": "You have 5+ years of experience with Python and PostgreSQL. Fully remote within European time zones. We can't sponsor visas. Salary EUR 75,000-90,000."
    }
  ]
}

If your data uses unusual field names, set Title field / Description field (e.g. headline, body).

Output example

{
  "title": "Senior Backend Engineer (Python)",
  "companyName": "Northwind Analytics",
  "location": "Remote (EU time zones)",
  "url": "https://example.com/jobs/northwind-backend",
  "id": null,
  "sourceDatasetId": null,
  "sourceIndex": 0,
  "seniority": "senior",
  "seniorityConfidence": 0.86,
  "jobFunction": "software_engineering",
  "jobFunctionConfidence": 0.93,
  "workMode": "remote",
  "workModeConfidence": 0.9,
  "remoteScope": "restricted",
  "remoteScopeConfidence": 0.71,
  "employmentType": "full_time",
  "employmentTypeConfidence": 0.62,
  "visaSponsorship": "not_offered",
  "visaSponsorshipConfidence": 0.88,
  "degreeRequirement": "not_mentioned",
  "degreeRequirementConfidence": 0.41,
  "postingLanguage": "en",
  "postingLanguageConfidence": 0.99,
  "managesPeople": false,
  "managesPeopleProbability": 0.08,
  "relocationSupport": false,
  "relocationSupportProbability": 0.03,
  "salaryStated": true,
  "salaryStatedProbability": 0.97,
  "minYearsExperience": 5,
  "lowConfidenceTags": ["degreeRequirement"],
  "descriptionShortened": false,
  "aiModel": "typesafe-ai/jev",
  "taggedAt": "2026-09-26T12:00:00Z"
}

Field	Values
title, companyName, location, url, id	Copied from your data (only these fields; the description is never copied to the output)
sourceDatasetId, sourceIndex	Where the posting came from — sourceIndex is its position in that dataset (or in items), starting at 0
seniority	intern, entry, mid, senior, lead (lead/staff/principal), manager, director_executive, not_clear
jobFunction	software_engineering, data_ai, it_infrastructure_security, product_management, design, sales_business_development, marketing_communications, customer_support_success, operations_project_management, finance_accounting, hr_recruiting, legal_compliance, hardware_engineering, science_research, healthcare, education_training, supply_chain_logistics, skilled_trades_production, hospitality_retail, administrative, other
workMode	remote, hybrid, onsite, not_stated
remoteScope	anywhere, restricted (specific countries, regions or time zones), not_stated; null unless the role is remote or hybrid
employmentType	full_time, part_time, contract (incl. freelance, temporary), internship, not_stated
visaSponsorship	offered, not_offered (incl. "must already have the right to work"), not_mentioned
degreeRequirement	not_mentioned, preferred_or_equivalent, bachelor_required, advanced_required
postingLanguage	en, de, fr, es, nl, it, pt, pl, other
managesPeople, relocationSupport, salaryStated	true/false, with the model's probability in …Probability
…Confidence	0–1: how sure the model is about the choice above
minYearsExperience	The smallest number of years of experience the posting asks for, read by rule (not by AI); null when no number is stated
lowConfidenceTags	Tags worth a human look: choice confidence below 0.5, or a yes/no probability between 0.35 and 0.65
descriptionShortened	true when a very long description was cut to its first ~4,000 and last ~1,800 characters before tagging; lines in between that mention pay, visa, relocation, work mode, degree or contract type are kept

A run summary (postings read, tagged, skipped, failed, datasets that couldn't be read) is stored in the run's key-value store as SUMMARY; postings that failed are listed in ERRORS.

Use it with

• Schedules: save your input as a Task and run it hourly, daily or weekly.
• Integrations: Slack, e-mail, Google Sheets and Google Drive, webhooks, Make, Zapier and n8n (the Integrations tab of the Actor or Task).
• API and code: the Apify API, the JavaScript and Python clients, or CSV/JSON/Excel downloads.
• AI agents: the Apify MCP server exposes this Actor as a tool; the input is plain text and short lists.

Pricing

Pay per result: $1.50 per 1,000 tagged job postings on the Apify Free plan; on paid plans $1.50 (Starter), $1.25 (Scale) and $1.00 (Business and higher) (the price for your plan is shown on the Pricing tab). The examples below use the Free-plan price. You only pay for postings that were tagged and saved — records without any text and postings that failed are never charged.

• Tagging one scraper run of 2,000 postings → $3.00.
• A daily feed of 500 new postings → about $22.50/month.

Set a maximum cost per run in the run options: the Actor checks it before each posting and stops cleanly when it is reached. On Apify's free plan a run tags up to 100 postings.

If you pick datasets in the form, the three sample postings prefilled under Or paste job postings are ignored (and not charged) unless you edit them.

How accurate is it?

We checked it on 24 public job postings from 9 employers (software, data, sales, support, compliance, fitness and gaming roles; US, Canada, Europe, India, Australia; one in French). We labelled each posting from its text first, then ran the tagger and compared. Where a posting itself is ambiguous (e.g. "work from a US hub or remotely"), either answer counted.

Tag	Matches
workMode	24 / 24
employmentType	24 / 24
visaSponsorship	24 / 24
salaryStated	24 / 24
relocationSupport	24 / 24
managesPeople	22 / 22
seniority	23 / 24
jobFunction	23 / 24
degreeRequirement	23 / 24
All tags	211 / 214 (98.6%)

The three misses: a "5+ years, AI product" engineer tagged lead instead of senior, a game-testing R&D engineer tagged data_ai, and a degree requirement read into a posting that names none. A sample this small is a sanity check, not a guarantee: keep an eye on lowConfidenceTags.

Limits and notes

• Tags come from TypeSafe AI's Jev, a decision model that picks from the fixed options above and reports its confidence; it never writes free text, so every value is one of the listed options. It can still pick the wrong option — use the confidence fields and lowConfidenceTags for anything important.
• English is the model's strongest language. German, French, Spanish and other languages work, but check a sample and lean on the confidence scores for non-English postings.
• minYearsExperience is read by rules (English and German phrasing such as "5+ years of experience", "mindestens 3 Jahre Berufserfahrung"). It takes the lower bound of a range and the smallest of several requirements.
• Postings are tagged several at a time. For very large datasets (tens of thousands of postings), raise the run timeout.
• The Actor doesn't visit any website: it only reads the datasets and postings you give it.
• If a long run is moved to another Apify server, it continues where it stopped: postings already tagged are not tagged or charged again.

FAQ

Is it legal to use? The Actor doesn't scrape anything; it classifies the job postings you provide, so you need the right to process that data (public job ads usually are). It outputs no personal data: e-mail addresses and phone numbers are removed from the text before it is analysed, and only the title, company, location, link and id are copied from your data.

Where is my data sent? Each posting's title, company, location, a few short structured fields (e.g. employment type, workplace type, tags) and its description are sent to TypeSafe AI's Jev model through an AI gateway (OpenRouter or Vercel AI Gateway) to compute the tags. TypeSafe states that it doesn't train its models on customer requests. Nothing else leaves Apify.

How accurate is it? Every tag carries the model's own confidence. When a posting is vague (no level in the title, no location policy), the model says so (not_clear, not_stated) or reports low confidence instead of guessing silently. Try it on 100 postings of your own data before tagging everything.

Can I tag postings from other job scrapers? Yes — any dataset of job postings works, as long as each record has a title and/or a description.

Can I get other tags? Open an issue with the tag you need; the question set is versioned in the changelog.

Can I use it from an AI agent? Yes, via the Apify MCP server or API. Pass items or datasetIds; the result is the dataset items above.

Support

Open an issue on the Actor's Issues tab with the input you used and the run link — we reply within a few days.

Made by Glidepath. See the changelog for updates.