Fields

A job has 50 fields: the GUIDs, the posting's own fields as its board states them, and the fields our models and rules add. A field that begins with ai_ is ours: a model's judgment, a rule, or a default for silence. The list derived on each job names the ai_ fields the posting did not state on that job, so a buyer that wants stated facts only can drop them.

Nothing is ever removed inside /v1. New fields are added as new names, each one a changed event for every job the day it is filled, and the changelog says when.

The job

Field	Type	Ours	Notes
`job_guid`	uuid		born when the job enters the feed, kept on reopen
`company_guid`	uuid		join to /v1/companies
`title`	text		the board's title, trimmed by rules; never null
`ai_title`	text	yes	the clean title; never null
`source`	text		the system (lever.co for Lever)
`url`	text		the employer's page
`description`	text		full text
`requirements`	text		full text, ours
`benefits`	text		full text, ours
`ai_workplace_type`	text	yes	On-site, Hybrid, Remote Solely, Remote OK
`ai_employment_type`	text	yes	FULL_TIME, PART_TIME, CONTRACTOR, INTERN
`ai_experience_level`	text	yes	years, as a text; null on about half the jobs
`ai_salary_min`	number	yes	yearly (hourly x 2,080), stated or read from the text
`ai_salary_max`	number	yes	yearly
`ai_salary_currency`	text	yes	defaults to USD
`salary_min`	number		as stated; null when none
`salary_max`	number		as stated
`salary_currency`	text		as stated
`salary_period`	text		HOUR, DAY, WEEK, MONTH or YEAR; null when none stated
`posted_date`	timestamp		never null (first seen when the board gives none)
`expiry_date`	timestamp		when the board states one
`date_modified`	timestamp		the board's
`first_seen`	timestamp		when our harvest first saw the job
`ai_job_language`	text	yes	English, French, ...
`ai_education`	text	yes	the level named
`ai_visa_sponsorship`	boolean	yes	false for silence
`location_json`	list		places: {city, state, country, lat, lng, precision, guessed}, town level
`remote_location_json`	list		places a remote job is open to; [] when none
`ai_languages`	list	yes	languages the posting asks for
`ai_onet_code`	text	yes	O*NET-SOC 2019; '' until filled
`ai_onet_title`	text	yes
`ai_soc_code`	text	yes	SOC 2018 (US); '' until filled
`ai_soc_title`	text	yes
`ai_noc_code`	text	yes	NOC 2021 (Canada); '' until filled
`ai_noc_title`	text	yes
`ai_isco_code`	text	yes	ISCO-08; '' until filled
`ai_isco_title`	text	yes
`ai_esco_code`	text	yes	ESCO; '' until filled
`ai_esco_title`	text	yes
`ai_summary`	text	yes	'' until the summary pass runs
`ai_skills`	list	yes	[] until filled
`ai_keywords`	list	yes	[] until filled
`countries`	list		what the country cut filters on
`timezones`	list		one IANA zone per place, in place order
`derived`	list		the ai_ fields the posting did not state on this job
`company`	text		the company's name; '' for a nameless board
`industry`	text		the company's industry, the short name
`is_active`	boolean		always true (open jobs only)
`sync_status`	text		always Active
`closing_since`	timestamp		set while a closure is held: keep the job out of matching

Types: text, number, boolean, timestamp (UTC, ISO 8601 with Z), list (a JSON array), uuid. A text that a model has not filled yet is "", never null; a list is [].

The company

A job's company and industry are its company's name and industry, denormalised; the rest of a company comes from /v1/companies by company_guid.

Field	Type	Notes
`company_guid`	uuid
`name`	text	'' when the board was imported with no name
`slug`	text	the board's key
`url`	text	the company's site
`size`	text	the employees band
`industry`	text	the short name
`locations`	list	a list
`ats`	text	the system
`ats_site`	text	the board's site
`board_name`	text	as the board names itself
`parent_company_guid`	uuid	the parent, when known
`slug_verified`	boolean	the slug was checked

Occupation codes

ai_onet_code and ai_onet_title (O*NET-SOC 2019, under CC BY 4.0; see Credits), ai_soc_code (SOC 2018, US), ai_noc_code (NOC 2021, Canada), ai_isco_code (ISCO-08, ILO) and ai_esco_code (ESCO, EU), each with its title. They are "" until the occupation pass fills them; the names are fixed now so nothing is renamed later.

Places

location_json is a list of places, each {city, state, country, lat, lng, precision, guessed}. Places are at town level, never a street address; lat and lng are the town centre from GeoNames (CC BY 4.0), null only for a place we could not match. timezones has one IANA zone per place in the same order; countries is the list the cut filters on. remote_location_json lists the places a remote job is open to, [] when none are stated.