Pepper Jobs Data API

How a record gets enriched

An ATS feed arrives messy and inconsistent per employer: years_experience carries “Regular/Limited/Flex”, education is “No” on 97% of jobs, pay hides in prose, and nothing states skills, benefits, visa policy or an occupation code. Closing that gap to one reference job shape is the product. Doing it without spending a model call on anything a lookup could answer is the engineering.

$1.10
total model spend, three organizations, 8,764 public jobs (pjd enrich report, 2026‑09‑18)
$157
the naive alternative for the same work: one request per job, full text, largest model
0%
escalation rate to the expensive tier, after the gates were tuned against the items they caught

The rule: the cheapest tier that can do the job

A model is never asked something a lookup can answer, and a larger model is never asked something a smaller one got right. A record enters at the top of the ladder. Each tier answers what it can, and those values leave for the enriched record; only the fields still null descend to the next, more expensive tier. Escalation is decided by code — the gates below — not by asking a model whether it feels sure.

one job record, as the ATS feed sends it the enriched record, as /v1 serves it SCHEMA MAP Sonnet · once per organization answers no field itself — it tells the rules below what each raw column of this feed really carries T0 · rules FREE · normalize.py pay · employment type · arrangement · locations · language T0 · strip FREE · enrich/prepare.py answers no field either — it removes this employer’s boilerplate before any model reads the text (−9% tokens) SCRAPE FREE · deterministic, no model schema.org JobPosting / Organization JSON-LD ONT-A · retrieval FREE · lexical, 61k O*NET titles the occupation code, when the title matches outright ONT-B · pick JEV · $0.042 / M input tokens the occupation code, chosen among ≤ 8 retrieved candidates T0.5 · classify JEV · one call per description job families · arrangement · experience · education · visa T1 · extract HAIKU · 6 descriptions per request requirements with evidence · skills · benefits · summaries T2 · escalate SONNET · one item per request only items a deterministic gate rejected — none so far down: the fields still null after this tier right: a value settled here, and it never descends again
The ladder, cheapest first. Two tiers answer no field of their own — the per-organization schema map and the boilerplate strip set up the tiers beneath them. The seven that do answer hand their values straight to the served record, so the work reaching Haiku is only what six free-or-near-free tiers could not settle, and the work reaching Sonnet is only what a deterministic gate rejected.
TierCostWhat it decidesWhere
SCHEMA MAPSonnetWhat each raw field really carries — once per organization, from a free statistical profile of the feed (~1K tokens)enrich/schemamap.py
T0 rulesfreePay strings, employment and arrangement labels, locations, languagenormalize.py
T0 stripfreeRemoves per-organization boilerplate before any model sees the textenrich/prepare.py
SCRAPEfreeschema.org JobPosting / Organization JSON-LD from the job and organization pagesscrape.py
ONT-AfreeTitle → O*NET-SOC by lexical retrieval over 61k O*NET titlesenrich/ontology.py
ONT-BJevPicks among ≤ 8 retrieved candidates, or none; low confidence defers to Haikuenrich/classify.py
T0.5 classifyJevJob families, work arrangement, experience, employment type, visa, education — one call per description, each answer with an explicit “not stated”enrich/classify.py
T1 extractHaikuDescription → structured fields, 6 descriptions per requestenrich/extract.py
T2 escalateSonnetOnly items that fail a deterministic gate; one item per requestenrich/extract.py

Step zero: the schema map, once per organization

Field names in the feed are a convention every implementation bends, and per-job extraction cannot fix it — the error is in what the column means. Measured across three organizations:

Raw fieldorg 2320 (demo)org 2089 (Workday)org 2012 (Oracle)
educationdegree text / “No”city namesemployment status (Regular / Temporary)
location_typeOn-site / RemoteOn-site Flexiblebusiness regions (“Japan”)
job_typeemptypay ranges, 100% filledParent Job / Child Job
brandemptyhospital divisionsaudience: Internal / External / Both
employment_typeFull-TimeFull time / Per diemフルタイム, Tiempo completo

So the question is asked once per organization, of a free statistical profile of the feed (pjd profile: top values per field, ~1K tokens), by the larger model, because it is a judgment call made once. The resulting field-role map steers the free rules for every job of that organization. Measured effect on org 2089: pay coverage 0% → 86% of 4,895 jobs, closing dates recovered from a custom field, and a 3% education rate removed that had been false positives from city names — for one request.

The map routes, it never invents

The map only routes raw fields to rules. The rules still validate every value, so a wrong map cannot invent data — at worst a field stays null. Two roles the map reports are never auto-applied: apply_url (the ones found were employee-only tenant links, not public apply URLs) and audience (which postings are public is an operator decision per source, and postings outside it never enter the API at all).

Why the bill is small: the cache is content-addressed

Model work is keyed by content, not by job. A re-synced job, a re-posted requisition, a second organization that happens to share a title — none of them costs anything, and neither does rebuilding the database, because the cache keys are derived from the text itself. Measured on org 2320 (2,700 jobs, ~3.2M description tokens):

The merge step is called by sync for every job, so a cached enrichment attaches the moment a job with a known description or title appears, and survives every re-sync.

The gates: what escalates, and what happens when nothing works

An item escalates from Haiku to Sonnet when any of these holds. All five are deterministic code (enrich/gates.py) — the model is never asked whether it is confident enough:

  1. It is missing from the micro-batch response, or fails schema validation.
  2. Pay is incoherent: min > max, no unit, or outside plausible bounds (HOUR 7–500, YEAR 10k–2M). Pay that merely differs from the feed is not a gate — measured, that was always a stated starting rate against the feed’s full range, and precedence already resolves it.
  3. A summary is empty, longer than 60 words, or a verbatim copy of a description sentence run over 25 words.
  4. ai_taxonomies_a is empty, or carries a value outside the enum.
  5. The model self-reports confidence: "low".

T2 output that still fails a gate is stored as failed and the job keeps its T0 values. Nothing is guessed to fill a hole.

Tuning the gates is free, and it moved escalation 31% → 0%

The first real run escalated 60 of 192 items (31%). Reading them showed that none was a model error: 51 were out-of-vocabulary taxonomy labels (“Nursing”, “Allied Health”) that only a transport without constrained decoding can produce — now dropped in cleanup, since a larger model is not the fix for a label typo; 7 were the text correctly stating a starting rate while the feed carried the full range; 2 were amounts with no unit, now settled by the same magnitude rule the rules apply to the feed. pjd enrich regate re-evaluated the stored results with zero model calls: 60 promoted, escalation 31% → 0%. Sonnet was never needed.

Every requirement is grounded, or it is gone

Structured requirements are the field most worth faking and the easiest to fake, so they are validated in code rather than trusted to the prompt. Each requirement carries evidence: the posting’s own phrase, copied verbatim. After extraction, ground() looks that phrase up in the stored description (case, whitespace and curly punctuation folded). A paraphrase or an invented requirement is not found and is dropped before storage — the count survives as ai_requirements_dropped, so the grounding rate is reportable rather than assumed.

Precedence, and the honest null

When two tiers have an opinion about the same field, the cheaper and more authoritative one wins. The order is fixed in code:

feed / rules  >  page (JSON-LD)  >  LLM  >  Jev

A structured statement from the employer beats an inference from prose. A Jev value only ever fills a field that everything above it left null, and its provenance records the confidence it was accepted at (jev:0.93). With include_beta=true every job carries an enrichment.fields map naming the tier behind each value, so the precedence is inspectable per record rather than promised in a document — that is exactly what the inspector shows.

A null is an answer

Every classification question can abstain, and every answer is thresholded. Jev always answers something; without an explicit “not stated” option it would name a work arrangement for the 22% of postings that say nothing about one. A code is never invented either: the ontology choice lists only retrieved candidates, and an answer outside that list is rejected in code. Thresholds are calibrated against answers that were already decided, not guessed — the occupation threshold sits at 0.75 because at that level Jev decided 77% of a 299-title calibration set with 91% agreement, and reading the 20 disagreements showed ~17 of them were the earlier label being wrong.

What it actually cost

Model work through 2026‑09‑18, three organizations, 8,764 public jobs, at list prices with no Batch discount (pjd enrich report):

StageModelRequestsItems$
schema mapclaude-sonnet-533 orgs0.03
ontology pickclaude-haiku-4-5781,936 titles0.28
extract (T1)claude-haiku-4-556274 descriptions + 55 shared paragraphs0.78
escalate (T2)claude-sonnet-5000.00
total1.10

Titles at that point: 1,512 resolved free (1,222 exact + 290 lexical), 1,853 by closed-list pick, 99 honestly “none fits”. Separately, the Jev tier took O*NET coverage from 70% to 86% for $0.07 on a 2,370-title backlog — Haiku’s estimate for the same backlog was $0.39 — and fleet-wide classification runs at roughly $0.17 per 1,000 descriptions.

Remaining work is quoted the same way: ~8,980 descriptions still to extract, estimated $18 via Batches. The prompt-version-2 re-queue of every description across five organizations and 10,410 active jobs estimates $24.85 via Batches, against a naive baseline of $273 for the same output.

What has not been validated

The agreement figures above are tier-against-tier, which bounds the error from one side only: two tiers can agree and both be wrong. Human review of the enriched fields is a separate, ongoing exercise, and the numbers it produces are kept apart from the ones here on purpose. See the inspector for the record-level view: source in, enriched out, tier per value.