\n Headquarters: Remote (US)\n
\n\nGong harnesses the power of AI to transform how revenue teams win. The Gong Revenue AI Operating System unifies data, insights, and workflows into a single, trusted system that observes, guides, and acts alongside the world’s most successful revenue teams. Powered by the G… [truncated for display]",
"workType": "remote",
"salary": null,
"employmentTypes": [
"Full-time"
],
"hiringContacts": []
}
```
Retained dedupe history can suppress previously returned postings. Keep a stable key and save your own history; an unchanged rerun can return zero new rows.
Recent AI and ML engineer jobs with complete original descriptions. Each row carries the source ID and URL, title, employer, locations, the original description as text and HTML, posting date, deadline, work arrangement, employment type, salary and the hiring contacts the source names, whenever the source publishes them. Unknown optional facts are `null`; an empty `locations` array means the source names no physical place, and an empty `hiringContacts` array means the source names nobody. `locations` is where the job is, `countryCodes` lists the countries of those places as ISO 3166-1 two-letter codes (`null` when no place has a known country), and `applicantLocations` is where an applicant may live, for example a remote job open only to residents of some countries: it is `null` when the source states nothing and `[]` when it states no restriction. `workType` is set only when the source names exactly one arrangement and is `null` when it names several, such as remote or hybrid; `locationTypes` matches a posting when any arrangement it names is selected.
Coverage includes these public job sources: Foorilla Hiring (formerly aijobs.net); Hacker News Who is hiring; Y Combinator's Work at a Startup; Built In; Remote OK, Remotive and We Work Remotely; Welcome to the Jungle; JustJoin.it. LinkedIn is also served, fetched live on every run. Each row keeps its own `source`. Every posting is verified before it is served. A posting with a missing, truncated or unverified description is never returned: every row has the full original text. LinkedIn results use the requested search filters. Each posting still has its full description or it is left out. LinkedIn may limit or block a run, so its rows can be fewer than requested or missing: the other boards are still returned, `RUN-SUMMARY.unavailableSources` names LinkedIn, and the run is marked `partial`.
Without `queries` (or the older `keyword`), it searches 19 default AI/ML phrases under Input and keeps a posting if its job title contains one. A company blurb that just mentions AI doesn't count anymore. We also keep postings with different titles, like AI platform or data engineering roles, if a separate model check of the public text clearly rates them as AI or ML work. That check runs every six hours, so new postings can show up a few hours later. Hacker News postings only use the title rule. For that check, the public text of postings is sent to an outside AI service; your searches and results never are. LinkedIn is the exception: it's searched live for six of those phrases and keeps its own relevance ranking. Your own `queries` still match in the title and description. Built In coverage is limited to the supported category, which is narrower than the whole site.
LinkedIn is searched live for at most 6 combinations of query, location and work arrangement per run. With the default `queries` it searches a curated subset of them, `computer vision`, `artificial intelligence`, `ml engineer`, `deep learning`, `machine learning` and `generative ai`, while the other boards use all of them. Naming several `locations` or `locationTypes` shares the six searches, so each location gets fewer of those phrases: two locations get the first three each, three get the first two, and four to six get the first phrase only. With your own `queries`, or more than six locations and work arrangements together, combinations beyond the sixth are not searched: the last locations you list can get no LinkedIn rows, the run is marked `partial`, and `RUN-SUMMARY.unavailableSources` says how many combinations were searched.
#### Input
```json
{"schemaVersion": "nomad-agent-simple-inventory-search-v2", "queries": ["machine learning", "deep learning", "artificial intelligence", "generative ai", "computer vision", "natural language processing", "reinforcement learning", "ml engineer", "data scientist", "data science", "applied scientist", "research scientist", "mlops", "llm", "nlp", "genai", "ai", "ml", "ia"], "locations": ["Europe"], "locationTypes": ["remote"], "employmentTypes": [], "postedWithin": "7d", "maxItems": 200, "maxItemsPerSource": 50, "dedupe": {"enabled": true, "key": ""}}
```
Every filter is optional, and an empty list means no filter. Filters are applied to all available postings before `maxItems` is counted.
- `queries`: a posting matches a query when its title or description contains every word of it, as whole words, ignoring case. A posting matches when any query matches. **Default:** a run that sets neither `queries` nor the older `keyword` searches for AI and ML engineer roles with these queries: `machine learning`, `deep learning`, `artificial intelligence`, `generative ai`, `computer vision`, `natural language processing`, `reinforcement learning`, `ml engineer`, `data scientist`, `data science`, `applied scientist`, `research scientist`, `mlops`, `llm`, `nlp`, `genai`, `ai`, `ml`, `ia`. On the stored boards it also keeps only postings whose job title contains one of them, or that a separate model check of the public text clearly rates as AI or ML work (Hacker News: title only). Your own `queries` or `keyword` replace them and are matched in the title and the description, and an empty list means no query filter.
- `locations`: a city, region, country name or two-letter ISO country code. Countries match by name, code and native name (`Germany`, `DE` and `Deutschland` find the same postings), and `EU` and `Europe` cover their member countries.
- `locationTypes` (`remote`, `hybrid`, `onsite`) and `employmentTypes` (`Full-time`, `Part-time`, `Contract`, `Internship`): while one is set, a posting whose source does not state that fact is left out.
- `postedWithin`: a positive duration in hours (`h`), days (`d`), weeks (`w`) or 30-day months (`m`), up to 36500 days, or `any`. Postings without a source date use the date they were first seen, which later refreshes never reset.
- `maxItems` defaults to 200; 0, or anything above 200, means 200.
- `maxItemsPerSource` (default 50, at most 200) caps how many postings each source contributes; `maxItems` then keeps the newest postings across all sources.
- `dedupe` suppresses postings already delivered to the same Apify account for the same search; different filters keep separate histories. Set `dedupe.enabled` to `false` to get stateless results, or set `dedupe.key` to keep separate alert histories.
#### Output
One flat row per posting in the `nomad-agent-job-row-v5` format: `schemaVersion`, `source`, `id`, `url`, `title`, `company`, `locations`, `countryCodes`, `applicantLocations`, `postedAt`, `deadline`, `description`, `descriptionHtml`, `workType`, `salary`, `employmentTypes`, `hiringContacts`. `employmentTypes` lists the types the source states (`Full-time`, `Part-time`, `Contract`, `Internship`) or is `null`. `countryCodes` lists the countries of the job's places as ISO 3166-1 alpha-2 codes (for example `DE`, `FR`), in the same order as `locations`, each once. A code comes from the source if it states one, otherwise from the place label. It's `null` when no place has a known country (a remote job with no country, or an unidentifiable place) and never an empty list. A place without a known country adds nothing, so the list can name fewer countries than there are places. A posting stored on a single-country board with no place at all carries that board's country, so `locations` can be `[]` while `countryCodes` names one. Postings read live from LinkedIn's public pages always have `countryCodes: null`. Those pages give a place label but no country code, so only stored postings carry a code. The `locations` filter matches countries against these codes as well as names and text. `applicantLocations` lists the regions where the source says an applicant may live (for example `United States` for a remote job open to US residents), `null` when it states nothing and `[]` when it states no restriction. Each hiring contact has a `name`, and a `title` and `url` or `null`. Newest postings come first. A run summary, including any input field that was mapped or ignored, is written to the `RUN-SUMMARY` record of the run's key-value store.
#### Changes
- **November 2026: We now include AI and ML roles under other titles.** The default search also keeps a stored posting if its title lacks the usual phrases but a separate check of its public text clearly rates it as AI or ML work. That check refreshes every six hours, so new postings can show up a few hours late. Hacker News postings still rely on titles only. Your custom `queries`, LinkedIn, and repeat-delivery history are unchanged.
- **November 2026: the default AI and ML search keeps AI and ML job titles.** Previously, a run with no specific search matched 14 phrases in either the title or description, so postings where the company text just mentioned AI would show up. Now it searches 19 phrases—including `ai`, `ml`, `genai`, and `ia`—and only keeps a posting if its job title contains one. Your custom `queries` still match both title and description, and LinkedIn (searched live for six of the default phrases) is unchanged. Using the default phrases starts a new repeat-delivery history, so postings you've already received might appear once more. Any task, schedule, or integration saved with the old 14 default phrases is now read as the default search.
- **October 2026: `countryCodes` added (row v5).** Rows are now `nomad-agent-job-row-v5`: one new field, `countryCodes`, lists the ISO 3166-1 alpha-2 country codes of the job's places, in the order of `locations`, each once (`null` when no place has a known country, never an empty list). A code is the country the source states, or if it doesn't state one, the country read from the place label. A place whose country can't be determined adds nothing. Postings read live from LinkedIn carry `null`; its public pages show a place label but no country code. `locations` and `applicantLocations` are unchanged. Input, other fields, and pricing are unchanged. Integrations validating against the v4 schema must allow the new field and the new `schemaVersion`.
- **October 2026: `applicantLocations` added (row v4).** Rows are now `nomad-agent-job-row-v4`: one new field, `applicantLocations`, lists where an applicant may live as the source states it (`null` when the source states nothing, `[]` when it states no restriction). `locations` is unchanged and still the physical place only; for a remote job open to some regions it is often `[]` while `applicantLocations` lists the regions. Input, other fields and pricing are unchanged. Integrations that validate rows against the v3 schema must allow the new field and the new `schemaVersion`.
- **October 2026: new row format.** Rows are now `nomad-agent-job-row-v3` (15 fields) instead of `nomad-agent-job-row-v1`. Postings come from the available postings of seven boards (Foorilla, Hacker News, Y Combinator, Built In, RemoteOK/Remotive/We Work Remotely, WTTJ and JustJoin.it), and LinkedIn is read live on every run. Pricing is unchanged.
- **Removed from every row:** `recordType`, `custom`, `warnings`, `docs`, `snippet` (the text is in `description` and `descriptionHtml`), `seniority`, `isRemote`, `remoteType` (use `workType`), `location` (use `locations`), `salaryMin`, `salaryMax`, `salaryCurrency` and `salaryPeriod` (the source's salary text stays in `salary`), `employmentType` (now the `employmentTypes` array, `null` when unknown), and `hiringContactName`, `hiringContactTitle` and `hiringContactUrl` (now `hiringContacts`, a list of `name`, `title` and `url`).
- **Input:** `queries`, `locations`, `locationTypes`, `employmentTypes`, `postedWithin` and `maxItemsPerSource` (default 50, was 36) filter the search, and `maxItems` defaults to 200 (was 288). `keyword` becomes `queries` and replaces the AI/ML default. `sources`, `cacheTtlSeconds`, `concurrency`, `runTimeoutSecs` and `apifyToken` are ignored with a note in `RUN-SUMMARY`. All supported boards remain included.
- **Repeats:** a posting you already received from this Actor for the same search is no longer returned again by default, where the earlier Actor only removed repeats within one run. Set `dedupe.enabled` to `false` for stateless results.
- **Source and ID changes:** postings from aijobs.net now carry `source: "foorilla"` (was `ai_jobs_net`). Remote board ids now start with the board name, for example `remoteok_123`, so a consumer that keys on `(source, id)` can receive postings it already has once more.
- **Board coverage:** Hacker News serves the current "Who is hiring?" thread only, and the full-text `searchQuery` mode is gone. Built In coverage is limited to the supported category. LinkedIn is fetched live, for a curated subset of the AI/ML phrases and at most 6 query, location and work arrangement combinations per run (see above), and can be limited or unavailable, in which case the run is `partial`; the other boards are still returned.
#### Fast setup and source code
[Public repository](https://github.com/Exdenta/nomad-agent-job-scrapers) · [Agent setup skill](https://github.com/Exdenta/OinkAIJobSearch/blob/main/.agents/skills/public-apify-actors/SKILL.md).
Use the Actor’s current input form for filters, or start with its documented API example. Select the `latest` build. Read the dataset together with the run summary: a successful status alone does not show that every source was reached.
Restrictive filters or repeat-delivery suppression can produce zero results.
##### Observed output example
Selected fields from a real source record inspected on 2026-10-03; consult the output schema for the full contract. Values change with the source.
```json
{
"schemaVersion": "nomad-agent-job-row-v3",
"source": "justjoinit",
"id": "accenture-industry-x---engineering-platform-integration-agile-product-implementation-analyst-consultant-gn--warszawa-pm",
"url": "https://justjoin.it/job-offer/accenture-industry-x---engineering-platform-integration-agile-product-implementation-analyst-consultant-gn--warszawa-pm",
"title": "Industry X – Engineering Platform Integration & Agile Product Implementation Analyst/Consultant (GN)",
"company": "Accenture",
"locations": [
"Warszawa, PL"
],
"postedAt": "2026-10-01",
"deadline": "2026-11-30",
"workType": "hybrid",
"salary": null,
"employmentTypes": [
"Full-time"
],
"hiringContacts": []
}
```
#### Product guides and related tools
- [Guide and starter input on nomadagent.dev](https://nomadagent.dev/actors/ml-ai-dev-bundle)
- [Product overview and public links](https://jobatlas.dev/actors/directory/source-jobs#nomad-agent-ml-ai-dev-bundle)
- [Public setup guide](https://github.com/Exdenta/jobatlas/blob/main/docs/public-actors-setup.md)
- [Job Atlas job-data tools](https://jobatlas.dev/actors)
- [Oink personal job alerts](https://oinkjobsearch.com/)
- [Search ten job boards with one API call](https://oinkjobsearch.com/articles/search-ten-job-boards-with-one-api-call/)
#### Current costs
Current Actor event rates, checked 2026-10-16. USD; use the **Pricing** tab for your plan and later changes.
| Event | Free | Bronze | Silver | Gold | Platinum | Diamond |
|---|---:|---:|---:|---:|---:|---:|
| Job result (`result`) | $0.0002 | $0.0002 | $0.00015 | $0.0001 | $0.0001 | $0.0001 |
| Actor start (`actor-start`) | $0.00005 | $0.00005 | $0.00005 | $0.00005 | $0.00005 | $0.00005 |
For example, 100 × `result`, 1 × `actor-start` totals **$0.02005 in Actor event charges on Free**. This is an arithmetic example, before any waiver; actual charged events depend on the run and selected options. It is not a promise of output or a charge for simply setting an item cap.
Your total is the charged Actor events, plus platform usage **only if this Actor’s Pricing tab bills it separately**, plus any external provider you use with your own key. Do not add platform usage twice when it is included. Apify credits and plan discounts affect your invoice. Set a run cost limit in Console before starting.
***
**Independent tool.** We built and run this Actor ourselves. The job boards and sites it reads do not operate, sponsor or endorse it, and their names are used only to say where the data comes from.
# Actor input Schema
## `queries` (type: `array`):
Match postings whose title or description contains every word of a query, as whole words, ignoring case. A posting matches when any query matches. Words match as the posting wrote them, so where postings are not in English, add the local term as a further query. Empty: no filter. Default: queries that find AI and ML engineer roles; replace them with your own, or clear the list to search every posting. With the defaults, a stored posting also needs one of those terms in its job title, or a high score from a separate AI and ML relevance check of its text (Hacker News: title only). Your custom queries match against both title and description.
## `locations` (type: `array`):
City, region, country name or two-letter ISO country code. Countries match by name, code or native name; EU and Europe expand to their countries. A posting also matches when the source names the place as somewhere an applicant may live (`applicantLocations`). Empty: no filter.
## `locationTypes` (type: `array`):
remote, hybrid or onsite. A posting matches when any arrangement it names is selected; `workType` in the row is set only when the source names exactly one. Postings whose arrangement the source does not state are left out. Empty: no filter.
## `employmentTypes` (type: `array`):
Full-time, Part-time, Contract or Internship. Postings whose type the source does not state are left out. Empty: no filter.
## `postedWithin` (type: `string`):
Positive duration in hours, days, weeks, or 30-day months, at most 36500 days; any disables the filter. Uses the first inventory admission date when the source has no posting date.
## `maxItems` (type: `integer`):
Maximum returned jobs. Zero, or anything above 200, means 200.
## `maxItemsPerSource` (type: `integer`):
Most postings any one source contributes, at most 200. maxItems then keeps the newest postings across sources.
## `dedupe` (type: `object`):
Suppress confirmed repeat deliveries for the same Apify user and search.
## `schemaVersion` (type: `string`):
Version marker for database-only simple job search. Version 1 inputs remain valid.
## Actor input object example
```json
{
"queries": [
"machine learning",
"deep learning",
"artificial intelligence",
"generative ai",
"computer vision",
"natural language processing",
"reinforcement learning",
"ml engineer",
"data scientist",
"data science",
"applied scientist",
"research scientist",
"mlops",
"llm",
"nlp",
"genai",
"ai",
"ml",
"ia"
],
"locations": [
"Europe"
],
"locationTypes": [],
"employmentTypes": [],
"postedWithin": "any",
"maxItems": 20,
"maxItemsPerSource": 5,
"dedupe": {
"enabled": true,
"key": ""
},
"schemaVersion": "nomad-agent-simple-inventory-search-v2"
}
```
# Actor output Schema
## `dataset` (type: `string`):
No description
## `runSummary` (type: `string`):
One record describing this run: what was attempted, what was delivered, and anything that degraded. Written to the run's key-value store under RUN-SUMMARY.
# API
You can run this Actor programmatically using our API. Below are code examples in JavaScript, Python, and CLI, as well as the OpenAPI specification and MCP server setup.
## JavaScript example
```javascript
import { ApifyClient } from 'apify-client';
// Initialize the ApifyClient with your Apify API token
// Replace the '