Why keyword search misses engineers
Keyword search misses engineers because a public profile holds only what the person chose to write about themselves. Across customer searches since June 2026, about half (~50%) of recruiter requirements cannot be answered from a public profile at all. The strongest engineers are usually employed and not applying, so you have to go and find them, and when you do, the profile alone rarely shows enough to judge them.
Skills inferred from peers, not guessed
Vort sourcing completes the skills an engineer left off their profile from what their actual peers do, rather than from a guess by a language model. The peers are people in the same role cluster at the same company, with at least 12 months of tenure. Every inferred skill shows how many of those peers it rests on, and its confidence falls in one of three bands: high (45% and above), medium (30% to 44%) or possible (under 30%).
An illustration from this site
This example is the Inferred Stack illustration on the Vort sourcing home page. It shows how the result is presented. It is not a customer search.
- The profile: a senior backend engineer, four years at the same company, 72 peers in the same role cluster.
- Listed on the profile: Python, AWS and Kubernetes.
- Inferred: Java, high confidence (47%), based on 23 of the 72 peers.
- Inferred: Spring Boot, medium confidence (38%), based on 18 of the 72 peers.
- Inferred: Kafka, possible (22%), based on 11 of the 72 peers.
Evidence from code, answers and talks
Vort sourcing reads six kinds of public signals across the web for each engineer: profiles and roles (titles, tenure, moves), code and projects (repos, packages, commits), answers and expertise (what they actually solve), company data (stack, scale, funding), writing and research (posts, essays, papers), and talks and communities (conferences, threads, posts). The engine cross-references them to infer the stack from the company and peers, estimate scale from infrastructure signals, weigh seniority against team size and read talks and filings for depth. Every inference shows its evidence: the source, and why the conclusion follows from it.
A second illustration on the Vort sourcing home page shows the result. The profile says only "backend engineer at a payments company". After reasoning from six public sources, the card shows a stack of Go, Kafka and Kubernetes, a real scale of about 40 million requests a day, a PCI-DSS and ledgers domain, a team of six that the engineer led, three talks and one patent, and a contact number. This too is an illustration, not a customer search. The candidates themselves come from public web sources, enriched with verified contact data, and nothing is scraped from behind a login wall.
A check against every must-have, with the receipt
Vort sourcing scores every engineer against every requirement in your brief, must-haves and nice-to-haves alike, and shows the receipt behind each score: a quote, a source and a live phone number. The home page illustration splits a backend brief into three must-haves (React, 5+ years of backend, fintech) and two nice-to-haves (Kafka, team lead), and counts the candidates who meet every must-have. In customer searches since June 2026, the median search screens 890 candidates and shortlists 16, with reasoning on every requirement.
Write the brief in plain language
You describe the engineer in plain language, the way you would explain the role to a colleague, with no Boolean strings and no filters to configure. A brief can be as short as the example in the home page search box: "Senior backend engineer · Tel Aviv · 5+ yrs · startup". Vort sourcing turns it into the query and the shortlist behind the scenes, and roughly ten relevant candidates come back in about 90 seconds, with verified contact details and ATS sync. You can then ask, in plain language, to re-rank, filter, compare or draft outreach, and the agent shows every step it took.
Vort is an AI sourcing platform that reasons over every public signal on a candidate: describe the role in plain language and get a ranked shortlist with a reason for every candidate and a verified phone or email. The full method is explained on how AI sourcing works. Who builds the product is on About, and questions on privacy, ATS integrations and data sources are answered in the FAQ.
- Can Vort sourcing find engineers who never updated their LinkedIn?
- Yes, that is what the inferred stack is for. It completes the skills an engineer left off their profile from what their peers in the same role and company actually do, and shows how many peers each skill rests on, so you can judge it yourself.
- Is an inferred skill just a guess by an AI model?
- No. An inferred skill rests on peers: people in the same role cluster at the same company, with at least 12 months of tenure. Each one is labelled high, medium or possible, with the number of peers behind it.
- Do I need Boolean strings to search for developers?
- No. You write the brief in plain language, and the query is built for you behind the scenes. Afterwards you can refine the results in plain language.
- How do I reach the engineers I find?
- Across customer usage since June 2026, a median 80% of candidates come with a verified phone or email in one click. Shortlists go straight into your ATS or out as a share link. Vort sourcing connects to almost every ATS, including CIVI, Adam Total, Niloosoft, HarmonyHR and VScout, as well as Greenhouse, Lever, Ashby, Workable, Comeet and Workday.
- Where does the candidate data come from?
- From public web sources, enriched with verified contact data. Nothing is scraped from behind a login wall.