SourcerDon't wait for applicants —
go find them
Surface hidden candidates that don't show up on LinkedIn — from conferences, papers, open source, patents, and community activity. We find them first. Personalized outreach drafts ready for you to review and send.
Job boards alone
can't build a strong candidate pool
Top talent in the AI era leaves verifiable traces across many public sources. Sourcer Agent gathers and evaluates those signals to build a strong candidate database.
Posting and waiting
only reaches candidates who are actively switching. The people you really need don't read job boards.
LinkedIn only tells half the story
Many engineers and researchers don't actively maintain a LinkedIn profile, or keep it private.
One recruiter can only do so much
Tracking hundreds of conference speakers, paper authors, and OSS contributors every week is hard for any human.
From one job post
to a pool of hundreds
We read your job, then cross-reference public sources to surface the people most likely to fit. The last word is always yours.
- 01Step 1
Data collection
Continuously scan conference talks, arXiv papers, GitHub contributions, patent DBs, and community activity. Public data only.
ConferencesPapersGitHubPatentsCommunities - 02Step 2
Fit analysis
Score candidates against your job's required skills, experience, and level — and produce the rationale for why this person.
Skill matchLevel estimationSwitch signals - 03Step 3
Personalized outreach
Drafts that reference the candidate's talks, papers, or projects in 1:1 messages. Nothing sends until you approve.
1:1 personalizationApproval gateFollow-up sequence
Three specialist agents
one workflow
Shadow Sourcer
Reconstruct candidates who aren't on LinkedIn (or barely so) using conference speakers, paper authors, open-source contributors, patent applicants, and community activity.
- Cross-validate across public sources
- Same-person inference across pseudonyms
- Profile evidence shown right inside the card
Fit Analyzer
Combine LinkedIn, GitHub, blog, conferences, and patents to score fit against the role and likelihood of switching — with the rationale for every recommendation.
- Match score against required skills and experience
- Switch likelihood from recent activity signals
- Reasons-to-reject captured too
Cold Outreach Agent
1:1 drafts that reference each candidate's talks, papers, or projects. Email → LinkedIn → follow-up sequences ready. Every send goes through the recruiter's approval gate.
- Per-candidate references inserted automatically
- Tone & language picker (Korean / English)
- Approval gate required to send
Where we find people
We combine public sources that are hard for any single person to monitor. We don't use anything that requires login or violates a service's scraping policy.
Local conferences
International conferences
arXiv / Papers
GitHub
Patent DBs
Tech communities
Workshops
Public portfolios
Nothing sends automatically.
The final call is human.
Sourcer finds candidates and prepares drafts, but the send button is always in your team's hands. Messages don't leave until they pass the approval gate. Every step can be reviewed so the outreach tone matches your culture.
- Auto-send
- 0%
- Human approval
- 100%
- PII retention
- Minimal
- AI
Candidate found
AI builds candidate cards from public sources
- AI
Fit score + rationale
Why this person, every time
- AI
Outreach draft
1:1 personalized, follow-ups included
- 4
Recruiter approval
HumanReview · edit · send is in your hands
Once you find them
what happens next?
Candidates Sourcer surfaces are scored by Probe against your company's bar. Each module runs on its own — but they're easier to use together inside Cofa's integrated HR Platform.
Evaluation & screening
Give every candidate the same AI environment for free, then auto-record their solving process into a reviewer-ready timeline.
Integrated system
Move surfaced candidates straight into the pipeline. Interviews, scorecards, and offers — the whole team in one screen, in real time.