> ## Documentation Index
> Fetch the complete documentation index at: https://docs.recruitier.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Job Search Overview
> Search thousands of jobs across the Netherlands with AI-powered semantic search.
## What Makes Recruitier's Job Search Different
Most recruitment platforms rely on simple keyword matching. You type "Python developer" and
only see listings that contain those exact words. Recruitier takes a fundamentally different
approach by combining **semantic understanding** with **keyword matching** to deliver results
that truly match what you are looking for.
When you search for "Python developer," Recruitier does not just look for those words. It
understands that this role involves backend engineering, scripting, frameworks like Django
and FastAPI, database experience, and related competencies. This means you also see relevant
listings titled "Backend Engineer," "Software Developer," or "Data Engineer" -- jobs that
a traditional keyword search would miss entirely.
Semantic search understands the **meaning** behind your query, not just the words. This
is why a search for "Frontend Developer" also surfaces jobs titled "UI Engineer" or
"React Developer" -- they describe the same role using different terminology.
## Three Search Modes
Recruitier offers three distinct search modes, each suited to different use cases. The
default mode is **Hybrid**, which gives you the best of both worlds.
| Mode | Description | Best For |
| -------------------- | --------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------- |
| **Hybrid** (Default) | Combines semantic understanding with keyword precision using Reciprocal Rank Fusion (RRF). Results are scored across both engines and merged. | General-purpose searching where you want both meaning-based and exact-term matching. |
| **Semantic Only** | Uses vector embeddings exclusively. Finds jobs based purely on meaning similarity. | Exploratory searches where you want to discover roles described with different terminology. |
| **Keyword Only** | Uses BM25 text matching exclusively. Finds jobs that contain your exact search terms. | Precision searches where you know the exact terms employers use, such as specific certifications or niche technologies. |
The Hybrid mode is recommended for the vast majority of searches. It ensures you do not
miss semantically relevant jobs while still rewarding exact keyword matches with a scoring
bonus. Switch to Semantic Only or Keyword Only when you have a specific reason to do so.
## How Jobs Are Sourced
Recruitier continuously scrapes job listings directly from company career pages across the
Netherlands. Unlike job boards that rely on employers to manually post listings, Recruitier
proactively discovers jobs at their source. This means you often find opportunities before
they appear on mainstream platforms.
The scraping pipeline works around the clock:
1. **Discovery** -- Recruitier's scrapers visit thousands of Dutch company websites and
career pages to identify new job postings.
2. **Extraction** -- Each listing is parsed to extract structured data: job title, location,
experience level, job type, flexibility options, salary information, and the full job
description.
3. **Enrichment** -- The platform links each job to its parent company record, pulling in
company details like industry, employee count, LinkedIn profile, and website.
4. **Indexing** -- Jobs are converted into vector embeddings and indexed for both semantic
and keyword search, making them instantly searchable.
Recruitier refreshes its job data continuously. New listings are typically indexed within
hours of being published on a company's career page. Jobs that are no longer available are
automatically marked as expired.
## The Search Database
The Recruitier search database contains tens of thousands of active job listings from Dutch
companies. Each listing includes:
* **Job title** and full description (with markdown formatting)
* **Location** with geocoded coordinates for radius-based filtering
* **Experience level** (junior, medior, senior, lead)
* **Job type** (full-time, part-time, contract, temporary, internship)
* **Flexibility** (remote, hybrid, on-site)
* **Salary range** when available
* **Company information** including name, industry, employee count, and LinkedIn profile
* **Posting date** for recency filtering (detected from multiple sources: JSON-LD structured
data, LLM extraction, HTML metadata, visible text patterns, URL patterns, sitemaps, or
the date Recruitier first discovered the listing)
* **Requirements summary** extracted by AI
* **Contact information** including emails, phone numbers, and LinkedIn profiles of decision
makers
* **"Geen acquisitie" detection** -- a flag indicating the company does not want recruiter
outreach
This rich structured data powers both the search filters and the AI-powered ranking that
makes Recruitier's results so precise.
## The Search Workflow
Using Recruitier's job search follows a clear, guided workflow:
Use the 4-step search wizard to define what you are looking for. Enter a job title,
review AI-extracted skills, set your filters, and configure preferences like salary
range and company size.
Recruitier processes your search asynchronously and presents a ranked list of matching
jobs. Each result includes a relevance score, job details, and company information.
Sort, filter, and explore the results to find the best opportunities.
Save the most promising jobs to your collection. Apply tags, add internal notes, and
track your outreach progress. Your saved searches are stored for easy access later.
Saved searches are automatically monitored. When new jobs matching your criteria are
posted, Recruitier notifies you and adds them to your search results so you never miss
an opportunity.
## Search vs. Saved Jobs
It is important to understand the distinction between **searches** and **saved jobs** in
Recruitier:
| Concept | Description |
| -------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Search** | A set of criteria (job title, skills, filters) that you define. Recruitier runs this against the database and returns ranked results. Searches can be re-run, cloned, and monitored for new matches. |
| **Saved Jobs** | Individual job listings that you have collected from search results or client pages. These live in your Jobs collection where you can tag, annotate, and track them. |
A single search can produce hundreds of results. You then selectively save the ones worth
pursuing into your Jobs collection for further action.
## Boolean Search Support
Recruitier supports boolean operators in the job title field to help you cast a wider net
for roles that could be described with different titles.
**How it works:**
* Use `or` between terms to search for multiple variations: `"python developer or backend engineer"`
* Each combination generates a separate search query to the recommendation engine
* Results from all queries are merged and deduplicated, so you never see the same job twice
* The final result set contains the best matches from all query variations
Boolean search with `or` is particularly useful for roles that go by multiple names.
For example, `"data engineer or data platform engineer or analytics engineer"` captures
all common titles for this type of role in a single search.
## Search Sources
Every search has a **source** that indicates where it originated. This helps you track
how your searches were created:
| Source | Description |
| ------------------ | --------------------------------------------------------------------------- |
| **Recommendation** | Created through the standard search wizard using the recommendation engine. |
| **LinkedIn** | Created from LinkedIn job data or candidate profile matching. |
| **Indeed** | Created from Indeed job data. |
The source is displayed on the search card in your saved searches list and on the search
detail page.
Follow the 4-step wizard to create a targeted job search.
Learn about the hybrid semantic and keyword engine behind the results.
## Key Features at a Glance
Enter a job title and AI automatically identifies the relevant skills, technologies,
and competencies to improve your search accuracy.
Combines vector-based semantic search with BM25 keyword matching for results that
balance meaning and precision.
Searches are processed asynchronously with live status updates so you always know
what stage your search is in.
Your searches are continuously monitored. New jobs matching your criteria are flagged
automatically, keeping your pipeline fresh.
## Advanced
### Search Entity Architecture
Under the hood, each search you create is stored as a **search entity** with a comprehensive
set of fields that drive the entire search pipeline:
* **Core identifiers**: name, description, and extra\_keywords that define what you are
looking for
* **Filter arrays**: experience\_level, job\_type, flexibility, and "when" (recency) are
stored as JSONB arrays, allowing multi-select for each filter dimension
* **Location data**: city name, latitude, longitude, and radius\_km for geocoded radius
filtering
* **Source tracking**: whether the search originated from LinkedIn, Indeed, or the
recommendation engine
* **Boolean type**: AND or OR logic for combining search terms
* **AI search flag**: marks whether AI-enhanced search features were used
* **Monitoring state**: is\_monitored tracks whether the search watches for new jobs, and
new\_jobs\_count tracks how many unreviewed matches have been found
* **Candidate linking**: candidate\_id connects a search to a specific candidate profile
for CV-based matching
* **Pagination**: pages controls how many pages of results to retrieve from the
recommendation engine
### How Search Connects to Other Features
The search system is the entry point to several downstream features:
1. **Job collection** -- Search results populate your personal job collection. Each result
becomes a Job entity linked to both the search and the underlying ScrapedJob record.
2. **Contact discovery** -- When results are returned, Recruitier enriches each job with
contact information from the parent company, enabling outreach directly from search
results.
3. **Candidate matching** -- Searches can be linked to a candidate profile. When a
candidate\_id is present, the search criteria are tailored to that candidate's skills
and experience, and results are presented in the context of that candidate's profile.
4. **Monitoring pipeline** -- Monitored searches feed into a background job that
periodically re-runs the search criteria against newly indexed jobs, delivering
real-time notifications via SSE (Server-Sent Events) when new matches are found.
5. **Cloning and collaboration** -- The search entity supports cloning to other users
within the same agency, copying all jobs, contacts, and metadata.
### Power-User Tips
If your initial search returns fewer results than expected, try these approaches:
* Use boolean OR to include alternative job titles
* Reduce the number of confirmed skills to the 3-4 most essential ones
* Widen the location radius or remove the location filter entirely
* Expand the "posted within" timeframe to 6 months or anytime
* Select multiple experience levels (e.g., both medior and senior)
You do not need to create a new search to refine results. Use the in-results filters
on the search detail page to narrow by experience level, job type, flexibility, match
type, tags, and active status. This is faster than re-running the search and preserves
your original result set.
Monitoring works best for stable, well-defined searches. If your search criteria are
very broad (e.g., just "Developer" with no filters), the monitor will find many new
matches and the new\_jobs\_count can grow quickly. For effective monitoring, use specific
job titles and at least 2-3 filters to keep the match stream focused and actionable.
### Business Logic Rules
* **Score normalization**: All match scores are displayed on a 30-99% scale. The top result
in any search is normalized to approximately 99%, with all other results scaled relative
to the top score. This means scores are meaningful within a single search but should not
be compared across different searches.
* **Poor match removal**: Jobs classified as "Poor Match" (matching less than 80% of
keywords) can be automatically removed to keep your results focused on quality
opportunities.
* **Deduplication**: When using boolean OR queries, each combination generates a separate
API call. Results are merged and deduplicated by the underlying ScrapedJob ID, so you
never see duplicate listings.
* **Search state progression**: Each search moves through states (Searching > Classifying >
Retrieving Company Info > Done) and cannot skip steps. Real-time progress is delivered
via SSE notifications.
## Related
* [Creating a Search](/jobs/search/creating-a-search) -- Step-by-step guide to the search wizard
* [How Search Technology Works](/jobs/search/how-search-works) -- Deep dive into the hybrid engine
* [Search Filters](/jobs/search/search-filters) -- Complete filter reference
* [Search Results](/jobs/search/search-results) -- Navigating and acting on your results