Hugging Face's logo

Datasets:
layerz
/
finance-engineer-postings

Modalities:
Tabular
Text
Formats:
csv
Languages:
English
Size:
< 1K
Tags:
finance
jobs
labor-market
fpa
ai-agents
job-postings
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
1
finance-engineer-postings / README.md
anthomakr's picture
anthomakr
Initial corpus: 37 finance engineering requisitions, 26 employers, frozen 2026-07-27
2bef7ad verified
preview code
|
Raw
History Blame Contribute Delete
4.98 kB
metadata 
license: cc-by-4.0
pretty_name: Finance Engineer Job Postings (2026 Corpus)
language:
  - en
size_categories:
  - n<1K
tags:
  - finance
  - jobs
  - labor-market
  - fpa
  - ai-agents
  - job-postings

Finance Engineer Job Postings (2026 Corpus)

Every job posting we could verify, in July 2026, for an engineering role sitting inside a company's own finance function. 37 requisitions across 26 employers, coded by hand from full text, with the sampling frame, the adjudication rules and the exclusion counts published alongside.

The role has no research literature. No occupational code, no salary survey, no professional body benchmark. Claims about it circulate as round numbers with no method attached. This dataset exists so the claims can be checked.

What is in it

finance-engineer-postings.csv, 37 rows, one per requisition.

Column Meaning
employer Company name as published on its own job board
ats greenhouse or ashby, the applicant tracking system the posting came from
department The department the employer itself assigns to the posting, verbatim
title Job title as posted
location Location as posted
published_date First publication date reported by the ATS
days_open_at_freeze Days between publication and the 2026-07-27 freeze
min_years_experience Minimum years required, when the posting states a number
tools_named Named tools, pipe separated (e.g. Python|SQL|dbt|Snowflake)
mentions_ai_agents Set when the posting references agents or LLMs
mentions_mcp Set when the posting names the Model Context Protocol
pay_band_usd Pay band exactly as posted, unparsed, currency and equity notes included
reporting_line Reporting line when the posting spells one out
posting_url Live posting URL at time of collection
archived_url Internet Archive snapshot, where the archiver was not blocked
collapsed_locations How many location variants of the same requisition were merged into this row

Blank means the posting did not state it, never that the value is zero.

How it was built

  1. A frozen convenience frame of 1,061 candidate employer board tokens was probed against the public Greenhouse and Ashby board APIs.
  2. 421 boards answered, carrying 27,278 open postings.
  3. 987 of those sat in a finance-side department.
  4. Postings were matched on the employer-declared department, not on the title. Title search does not work here: all 37 titles are distinct as posted, and 16 contain no finance word at all.
  5. Raw matches were adjudicated by hand against four written exclusion rules, then multi-location duplicates were collapsed. 97 raw matches, 58 excluded, 39 postings, 37 requisitions.
  6. Survivors were coded from full text and snapshotted.

Full method, adjudication rules R1 to R4 with their exclusion counts, and the analysis: https://app.layerz.cc/resources/finance-engineer-statistics

Limitations, stated plainly

  • The frame is a convenience sample, skewed toward US technology employers that publish on Greenhouse or Ashby. It is not a census and it does not support national estimates.
  • Only two ATS platforms. Workday, Lever, SmartRecruiters and in-house boards are absent.
  • Live postings only. A corpus of currently-open roles is recency-biased by construction, so the age distribution says less about the role's novelty than it appears to.
  • The R1 boundary is the weak point. At fintechs, "builds financial products for customers" and "builds the finance function's systems" overlap. Several employers produced matches excluded on that basis. The rule is written down so you can disagree and put them back.
  • pay_band_usd is raw text, not normalised. Parse it yourself; a naive regex on decimal bands gets it wrong.

Reproducing or extending it

The scripts, the frozen frame and a runbook covering the pitfalls are published with the method. The corpus is refreshed quarterly against the same frozen frame, so counts stay comparable across runs. Next run around October 2026.

If you extend the frame, report the extension separately rather than merging it, otherwise a rise in the count comes from the frame rather than from the market.

Citation

Corpus frozen 2026-07-27. CC BY 4.0, reuse with attribution.

Layerz (2026). The Finance Engineer Role: Job Posting Statistics. https://app.layerz.cc/resources/finance-engineer-statistics

Related

Compiled by Layerz.