(Internal) AI sourcing: How to write filters

🔍 Articles in This Section

Please use the following list to see additional internal articles regarding AI Sourcing:

• AI Sourcing: Overview
• AI Sourcing: How to write filters

General guidance

• Gem’s filters work similarly to LinkedIn Recruiter search
• The goal of filters is to carve out a “talent pool” so the AI can rank relevant candidates and give you the best ones
• Make sure your talent pool is big enough.
• Ideally you want 5k - 20k candidates in the talent pool
• A pool of <1000 for a non-niche role may mean your filters are too specific (ormisconfigured)
• A common mistake is going too narrow, and winding up with only a few hundred candidates (or less!)
• Make sure your talent pool isn’t too broad.
• Again, target a pool size of 5k - 20k
• Use your intuition…
• a pool of more than 20k for a non-niche, location-specific role
• or a pool of > 30-40k for a non-niche, location-agnostic remote role
• …could mean your filters are too broad.
• When a pool is too large, Gem AI can only score a fraction of the pool — so unless everyone in the pool is truly a potential fit, you won’t get as good of results
• Most filters are “contains” filters by default - these work by substring matching
• This means that you can use keywords to do broader searches - for example “software” will match all of “Software Engineer”, “Software Developer”, and “Software Development Engineer”
• This style is similar to “boolean searches” on LinkedIn, except that it’s for just one field (title / school / company / etc) and we only support “OR”

Specific filters

Job title

• Job titles vary a lot between companies, even for the exact same role
• “software engineer” or “product engineer” or “API developer”?
• “enterprise architect” or “solutions engineer” or “sales engineer” or “implementation manager”
• “technical support engineer” or “customer support analyst”
• when building a search, include many title variants
• research which titles are common for your role (and similar ones), and be sure to include all of them in the search to widen the talent pool
• use keywords to hit many titles at once
• “software” can hit lots of engineer titles; “support” can get lots of support roles
• it’s better to include some irrelevant candidates than to exclude large swaths of qualified candidates — the AI step will find you the best matches
• think about current vs current & past
• how important is it that candidates are currently in this role? Would someone who transitioned away potentially want to come back to it?

Location

Country / State / Metro / City

• We have 4 location filters, but they’re a hierarchy: Country > State > Metro > City
• You probably only need to use one of the four filters
• If you enter “city=San Francisco”, you don’t need to enter anything in the other filter
• entering SF implicitly filters on “metro = San Francisco Bay Area”, “State = California”, and “Country = United States”
• The 4 location filters are AND’ed together, which may give unexpected results if you use more than one
• if you enter “City=San Francisco” and “State=Texas” you will get no results, because no one is located in “San Francisco AND Texas”
• Within a given filter you can enter more than one value, and we’ll OR them
• you can use “San Francisco” OR “Austin” to get a talent pool with candidates in both cities
• Remember to go wide when possible
• Overly strict location filters can shrink your talent pool and give poorer results
• If your role is in-office in SF, consider at least “San Francisco Bay Area” since folks may be willing to commute. Or even consider all of CA (or other states), if you offer relocation / think candidates would be willing to move.

Experience & Tenure

Months at current company

• Straightforward, months at the current company (in any role)

Overall years of experience

• This is total years of work experience in any role.
• This includes internships and anything else listed in the LinkedIn profile as work experience
• This filter is not “years of experience as a

” or “full-time experience excluding internships” - those work better as qualifications for the AI to evaluate

• It can be helpful to pad out your experience requirement by a few years when using this filter, because of internships / non-relevant experience / etc

Companies

Companies

• this is currently immediate matches only - we don’t have “similar companies”
• this filter is a common way to overly restrict the talent pool
• if you enter companies here, you’ll only get candidates that worked at exactly these companies
• for now, best guidance is probably to leave this blank
• if you do use it, be sure to include a very wide list of companies
• also be sure to uncheck “current company only” to broaden the search

Education

School

• Similar to companies, this is immediate matches only
• only use this if you only want candidates who specifically attended just this list of schools
• this filter can shrink your talent pool by a lot

Degree

• this field is not yet normalized - so there are tons of variations
• e.g. “BS”, “Bachelor’s Degree” “Bachelor’s Degree (B.S.)”
• use keywords to get around the variation
• e.g. “bachelors” OR “bs” OR “b.s.” will cover most bachelors degrees
• unless degree is very important (e.g.medical / law degree) consider leaving this blank
• because of the variation it’s easy for this filter to backfire and severely limit your talent pool
• most jobs require “bachelor’s degree or equivalent experience”, and usually job title filters alone give a talent pool that meets this requirement

Field of Study

• schools have many different names for the same major / field of study
• Computer Science also may be called “Software Engineering” or “EECS / Electrical Engineering & Computer Science” etc
• use keywords to target a wider set of fields
• for example “engineering” OR “science” would cover most technical fields
• it’s better to be too wide than too narrow here - you can use AI qualifications to narrow in
• unless field of study is critically important, consider leaving it blank
• work experience is often more important as a hard filter
• variations in this field can cause this filter to limit your search

• General guidance
• Specific filters
• Job title
• Location
• Experience & Tenure
• Companies
• Education