Define core criteria for a job

Why criteria matter

This is the step where you tell Talinty's AI what "a good fit for this role" actually means. Skills, experience, education,

languages, location, certifications. The AI reads every resume, every assessment, every interview transcript against the

criteria you set here.

Vague criteria produce vague scores. "Strong communication skills" doesn't help; "professional written communication in

English and French" does. The model is precise when you are.

Criteria are editable any time. If you discover mid-search that the requirements should be tighter (or looser), update them; the

pool re-ranks automatically. No re-screening required.

The criteria categories

The wizard prompts you through a standard set of categories. You don't have to fill in all of them; fill in the ones that matter

for this role.

Skills. Technical (Python, SQL, AWS), soft (project management, customer-facing communication), or domain-specific

(regulatory knowledge, language fluency). Add each skill as a separate entry and tag it Required or Preferred.

Experience. Years in the role, years in the industry, specific past job titles. You can set a minimum, a preferred range, and a

maximum (the maximum is useful for early-career roles where over-qualified candidates would likely leave).

Education. Degree level, field of study, specific institutions. Tag each as Required, Preferred, or Equivalent (where equivalent

experience is acceptable in place of the credential).

Languages. Required and preferred languages, with fluency level (Basic, Conversational, Professional, Native). For roles in

MENA, both Arabic and English are commonly required.

Location or availability. Time zones the role must overlap with, in-person attendance requirements, willingness to relocate.

Certifications. Industry-specific credentials (PMP, AWS Solutions Architect, CFA).

Required vs Preferred

Every criterion is tagged one of two ways:

Required. A candidate without this criterion gets a substantial penalty in their AI score. Use sparingly. Three or four required

criteria is usually enough; more than that and you're filtering everyone out.

Preferred. A candidate without this criterion isn't penalized but doesn't get the boost either. Most criteria should be Preferred

unless they're genuinely deal-breaking.

The AI doesn't auto-reject candidates who miss required criteria. It scores them lower and flags the gaps in the Things to

consider column of the vetting brief. You decide whether the gap is disqualifying.

Weights, for power users

By default, each criterion contributes proportionally to the AI score. If you want one criterion to count more than the others

(for example, "this role lives or dies on Python depth"), you can adjust its weight from 1x (default) to 3x.

Weight adjustments are most useful for senior or specialist roles where one capability dominates the decision. For most

generalist roles, default weights work fine.

Deeper AI tuning (recommendation thresholds, score calibration, troubleshooting outputs) lives in the Tuning the AI category.

This article keeps the basics every recruiter needs; the AI category covers the advanced moves.

The "too narrow vs too broad" trap

Two common mistakes:

Criteria too narrow. "10+ years of Rust experience, PhD in distributed systems, native in Mandarin, based in Lyon." Three

candidates in the world match. You'll get four applicants and reject them all.

Criteria too broad. "Software engineer, some relevant experience, good communicator." 800 applicants. The AI can't tell

which ones to surface; you have to read everyone.

The sweet spot: criteria specific enough that the AI can rank meaningfully, broad enough that the pool is bigger than 20

candidates. If you're getting an obviously wrong distribution after the first day of applications, edit the criteria and let the pool

re-rank.

Editing criteria mid-search

You don't have to get this perfect on day one. Editing criteria after the job is published has three behaviors worth knowing:

The AI re-scores every existing candidate against the updated criteria. Scores can go up, down, or stay the same

depending on what changed.

The vetting briefs update to reflect the new criteria. Why they match and Things to consider both rewrite.

The pipeline view re-ranks. A candidate who was at the top might drop; one in the middle might rise.

The recompute is fast (usually under a minute even for large pools) and reversible. If you don't like the new ranking, revert the

criteria edit and everything goes back.