Candidate Sourcing: 17 Sources for Hiring Niche Experts

The best specialist for your vacancy does not respond to job ads. They have a job, they are happy in it, and their resume on hh.kz has most likely not been updated in ages. If you wait for applications, you will hear from people who are looking for work right now. That is why, for expert roles, recruiters usually have to go and find people themselves. Candidate sourcing means finding people who fit the role but have not applied yet: through resume databases, professional sites and public profiles. Below we explain how this search works in HRStream: what the 17 sources are, what each gives a recruiter, and which pain point it solves.

Why one resume database is not enough for expert hiring

For high-volume roles, a resume database and incoming applications work well: there are plenty of people, and they want the job. With a rare specialist it is the other way around. According to a LinkedIn survey, retold in January 2026 by Staffing Industry Analysts, 66% of recruiters say finding qualified candidates has become harder (6,554 HR professionals worldwide, countries are not broken out in the report). And in a SHRM survey, 77% of U.S. organizations admitted they find it hard to hire for full-time positions. This is not Kazakhstan data, but the picture will be familiar to local recruiters.

Manual sourcing also comes with five typical pain points:

• Many channels, no time. A recruiter opens LinkedIn, then hh.kz, then GitHub, then Behance, and searches from scratch in each, phrasing the query differently every time.
• The same person lives in several places. They get found twice and contacted twice.
• The query has to be rewritten for every platform. Job titles, synonyms and filters are different everywhere.
• It is unclear what to do when too few people are found. The decision to "relax the requirements" is made by gut feeling.
• What you find stays in bookmarks. People from search results have to be moved into the hiring system by hand.

Cold search in HRStream solves these five problems in one window. It is available to the organization's owner, administrator and recruiter. It is closed to clients and observers.

What the search looks like: from vacancy description to a list of people

You write the query the way you would explain it to a colleague. For example: "Economist in Aktau, 1C and IFRS", "Certified welder (NAKS), Atyrau, rotational shifts" or "Python developer, remote, 3+ years". HRStream splits it into clear parts: job title, similar titles, location, must-have requirements, nice-to-have requirements and conditions (schedule, work format, experience). The breakdown is visible on screen, and you can correct any part. If the automatic breakdown is unavailable, you set the conditions manually, and the search works the same way.

Must-have and nice-to-have skills do not screen people out, they affect the order: the better a candidate matches them, the higher they appear in the list. So a strong specialist will not drop out just because their profile lacks one word from your query.

The same query goes to all the sources you selected. You do not have to rewrite it for each platform: HRStream adapts the wording itself. A source that cannot apply a particular condition says so plainly: "Not applied: …". It does not widen the search silently, so you always know which conditions the people were selected by.

17 sources: what each one gives you and when you need it

The sources fall into three groups: your own data and resume databases, professional platforms, and social networks. Each group has its own role.

Group 1. Resume databases: where people look for work

1. HRStream's own database. Everyone who has ever applied to your vacancies, came from your channels or was added by a recruiter. The search covers the desired position and the job titles in work experience. Pain point it solves: you pay to attract candidates and then forget about them. All the selection filters work here, and for a saved search you can switch on a "N new candidates" notification. Such notifications exist only for searching your own database.

2. HeadHunter. Resume search on hh.kz, hh.ru and hh.uz (you need to connect HeadHunter; how to do it is covered in a separate article). Results come as cards without contact details. The full resume is loaded for the best candidates. Pain point: no need to sit in the HH account and copy resumes one by one. The candidates you choose land in HRStream. Contact details are revealed by a separate action with confirmation and use up your HeadHunter account's daily limit.

3. Avito. Search across Avito resumes: by job title, city and experience. Results are anonymous too. If the candidate has applied to your vacancy, contact details open at no charge. In other cases this uses up your Avito package, and without an active package the contact will not be revealed and nothing will be charged. Avito does not use job title synonyms, and cards here are evaluated manually. More about connecting it in the article on the Avito integration.

Group 2. Professional platforms: where experts show their work

This is where you find people who do not update their resumes on job sites but publish code, design and solutions. HRStream shows public information: what the person chose to share about themselves.

• 4. LinkedIn. Name, job title, company, location, education, experience and number of connections. This is the main source for almost any role, and it is always on. Pain point: you need to find executives, managers and specialists who do not use local job sites.
• 5. Habr Career. Specialization, job title, grade and the "looking for work" status. Pain point: telling apart those who are open to offers from those who are not.
• 6. GitHub. Company, location, link to a blog or portfolio, profile description, an "open to offers" flag and number of repositories. Pain point: seeing real code instead of words in a resume. If the person has listed a public e-mail themselves, it may be visible in the profile details.
• 7. Stack Overflow. Name, location and "about me" text. Pain point: finding someone who helps other developers and knows niche technologies.
• 8. Kaggle. Job title, company, location and rank. Switched on for data science and machine learning, and for keywords such as "neural networks". Pain point: data specialists are almost absent from classic job sites.
• 9. Behance. Role, studio, location and an "available for hire or freelance" mark. Pain point: a designer should be judged by their portfolio, not by text.
• 10. Dribbble. Name, role and location. The platform does not allow storing copies of its results, so every page is requested anew. Pain point: finding strong product and interface designers.

Group 3. Social networks: when the expert is not on professional sites

Sometimes the person you need is a blogger, a craftsperson, a photographer or a salesperson, and they show their work precisely on social networks. HRStream has a strict mode for them: only the name, the link and the city (if it is explicitly stated) are shown. Bios, photos, age, family and personal details are not collected. The card carries a direct warning: evaluate professional information only.

• 11. YouTube. Channel name and link. Suitable for hosts, videographers, teachers and authors of educational content.
• 12. Instagram. Profile name and a profession from a limited list. Switched on for media, marketing and creative roles.
• 13. TikTok. Username and link. This is an experimental source: use it as a supplement.
• 14. VKontakte. Name, link and city. Switched on automatically for blue-collar and skilled trades, and for other roles at your choice.
• 15. Facebook, 16. Odnoklassniki, 17. X. Name, link and city. They are not switched on automatically for any role, and you can connect them manually when you need to test a rare hypothesis. For most roles they give few results, and the system warns you about it.

Telegram is not among these 17: you cannot search for people on it. But if a person has listed their Telegram in another profile, HRStream recognizes that link and uses it to match profiles.

How HRStream picks sources for a role

You do not need to know where to look for a designer and where for a welder. HRStream determines the role type from the job title and switches on the suitable sources:

If no suitable source is found for a non-standard role, HRStream connects the closest ones, because an empty result is worse than a weak one. Any source can be switched on or off manually. The organization's administrator can close individual sources for their team altogether.

What to do when specialists are scarce

For expert positions, "4 people found" is a normal situation. The search does not leave you alone with it.

• Exact matches first, then synonyms. The specialized search exhausts the exact job title and only then moves on to similar titles, neighboring cities, adjacent roles and the whole country.
• Expansion suggestions. If few people are found, HRStream suggests switching on, one by one, adjacent roles, dropping the experience requirement, dropping the schedule and work format, or searching the whole region. Each step can be undone with the "Undo expansion" button.
• "Show more" and "Search deeper". The first batch can be extended. For a nationwide search, the system warns that the city will not be taken into account.
• One person, one card. Profiles of the same specialist from different platforms are merged. If the match is doubtful, HRStream writes "possibly the same person" and does not merge cards at random. Those already in your database get an "Already in database" tag.

From list to funnel: evaluation and moving to a vacancy

Candidate evaluation. For your own database and HeadHunter, HRStream evaluates cards against the candidate profile: first quickly from the card, then in full from the complete resume for the best ones. Criteria that cannot be checked from the card do not lower the score. The shortlist is built from your marks and the selected candidates, and the list can be sorted by relevance or by score. Profiles from LinkedIn, GitHub and other professional platforms are not yet evaluated automatically and do not enter the score-based shortlist: you review them yourself. On Avito, evaluation has to be done manually.

Moving to a vacancy. You add the people you need to the funnel with the "To vacancy" button. The candidate lands at the chosen stage as a cold contact. For profiles, a card without contact details is created, and duplicates are detected automatically. The move itself does not use up contact reveal limits. What happens next is described in the article on AI candidate screening.

Working as a team. Cold search is available to the owner, administrator and recruiter roles, and how roles and analytics work is covered in the article on managing a recruiting team.

Personal data: what to know before the first message

A public profile is not permission to do anything you like with the data. The Law of the Republic of Kazakhstan "On Personal Data and Their Protection" distinguishes between publicly available data and restricted-access data and attaches importance to the subject's consent. This is not legal advice: for how to store and use the contact details of people you find, consult a lawyer or your company's data processing policy.

The product itself includes several safeguards:

• age, date of birth, marital status, phone numbers and photos from profiles do not make it into cards; a public GitHub e-mail, if the person listed it themselves, may be visible in the profile details;
• data from open sources is marked as preliminary and carries the date it was obtained;
• personal social networks carry a warning that personal attributes must not be used in evaluation;
• age, gender, marital status and other protected attributes in the free text of a query are automatically removed from the search and from the evaluation;
• a recruiter can submit a request to hide a profile at the owner's request: the profile disappears from results for all organizations, and only an anonymized trace of the request is stored in the system.

A practical rule for the recruiter: collect the minimum, name the source and purpose in your first message, and delete data at the person's request.

In practice: how to use sourcing for an expert position

1. Break the query down first. Separate the role from the requirements: the job title drives the search, the skills drive the order.
2. Start with automatic source selection. Add the others only if there are few results.
3. Look at what the system did not take into account. The "Not applied: …" line shows what is worth checking manually.
4. Expand the search one step at a time. Adjacent roles first, then experience, then region.
5. Reveal contacts only for those who passed your screening. HeadHunter and Avito limits are finite, and revealing is a separate confirmed action.
6. Move people to the vacancy right away. Do not copy links into notes: the people you find should land in the funnel, where the whole team works with them.

How HRStream solves this

Cold search in HRStream is a single window for finding candidates across 17 sources. It understands a natural-language query, picks sources for the role on its own, suggests how to widen the search, does not merge people at random, evaluates candidates from your database and HeadHunter against the profile, and moves the right ones into the vacancy funnel. Revealing contact details remains your explicit decision. For recruitment agencies, this search is convenient for closing clients' expert orders: details are in the article on an ATS for recruitment agencies.

Frequently asked questions