Adopt a Dataset

šŸ¾ Adopt a Dataset

I'm looking for a good home for some government datasets.

šŸ” A good home

I put some work into getting these datasets ready, but I donā€™t have time to tend to the care and feeding of all of them. Iā€™m sure Iā€™ve made mistakes, and theyā€™re hard for me to catch because Iā€™m not a user of most of these datasets. Iā€™m looking for someone who wants to claim each one: maintain it, give it regular attention, keep it open and available to the community, and reach out to the researchers and others who might use it.

  • All of these run automatically and pull data from a government website. Most have no API that the government hosts, and many involve PDFs or web scraping. On Hugging Face you can query all of them in one place, and the websites let anyone filter and find what they need.
  • Each one comes with a GitHub repo for the code, a Hugging Face repo for the data and a basic website. Iā€™m hosting the websites for now, but Iā€™d like to stop.
  • To take one over, you need an engineer or someone engineer-like. Someone who will respond when something breaks, check on it now and then, and fix the code when something is off.
  • Some need an LLM API key, because I pull fields out of unstructured text. I use Claude. The bills are small (see each card). Thereā€™s no other spend.
  • I wrote a handoff document for each one to help your engineer take over. Youā€™ll need to sign up for a few free accounts.

Meet them

Federal J&A Documents

Why federal contracts skipped competition.

  • Refreshes daily
  • Live sites
  • LLM key
Size
20,019 notices and 22,160 attachments (about 11 GB)
LLM cost
About $4.50 a month

What you get

  • Every justification notice on SAM.gov since January 2025, with the original files and text
  • An LLM pulls out the vendor, dollar value, legal reason and a summary (experimental)
  • A daily email digest of the notable ones

Code: Apache 2.0 Ā· Data: CC0

DatasetWebsiteDaily digestCodeHandoff guide

OTA Watch

Federal ā€œOther Transactionā€ deals, made outside the usual contracting rules.

  • Refreshes daily
  • Live site
  • LLM key
Size
59,581 records, 2000 to 2026
LLM cost
About $1 a month

What you get

  • Every Other Transaction record from FPDS, the federal contracting database: agency, vendor, dates and dollars obligated
  • A flag for deals run through a consortium

Code: MIT Ā· Data: see dataset card

DatasetWebsiteCodeHandoff guide

DoD Daily Contracts

Defense contract awards, one row each.

  • Refreshes daily
  • Live site
Size
43,612 awards since July 2014

What you get

  • Every contract award the Defense Department has announced: date, branch and full award text
  • The company and its city and state, where they could be read

Code: MIT Ā· Data: public domain (US government works)

DatasetWebsiteCodeHandoff guide

EEOC Appellate Decisions

EEOC decisions on federal employeesā€™ appeals.

  • Refreshes daily
  • Live site
  • LLM key
Size
115,882 files, about 5 GB
LLM cost
About $0.003 per new decision

What you get

  • Every federal-sector appeal decision EEOC posts, with the original files
  • An LLM pulls case details from decisions since 2025: the agency, grounds claimed and outcome (5,744 so far)

Code: MIT Ā· Data: public records

DatasetWebsiteCodeHandoff guide

OPM Union Contracts

Every federal union contract, as PDFs and text.

  • Refreshes weekly
  • Live site
Size
1,248 contracts

What you get

  • The PDFs with their text, plus the agency, union and dates OPM lists
  • A first-pass flag for contracts terminated under the 2025 executive orders

Code: MIT Ā· Data: OPM-published documents, unofficial copy

DatasetWebsiteCodeHandoff guide

HHS Appeals Board Decisions

HHS Appeals Board decisions back to 1974.

  • Refreshes daily
  • Live site
Size
9,853 decisions, plus 224 Medicare Appeals Council decisions

What you get

  • Appeals decisions back to 1974 and administrative law judge decisions back to 1985, with full text
  • Case details from each decisionā€™s header

The original corpus was built by KMisener90 (this repo is a fork, MIT). The live site and dataset are mine.

Code: MIT Ā· Data: US government works

DatasetWebsiteCodeHandoff guide

DC Procurement

DC procurement records, kept after the portal drops them.

  • Refreshes weekly
  • Live site
Size
885,916 payments, 243,611 purchase orders and 7,544 contract periods

What you get

  • DCā€™s contracts, purchase orders, payments and solicitations
  • Weekly history, so records the portal drops stay

Code: MIT Ā· Data: see dataset card

DatasetWebsiteCodeHandoff guide

Contract Terminations

Federal contract terminations, with the reason.

  • Refreshes monthly
  • Live site
Size
67,966 terminations since October 2024

What you get

  • The reason (convenience, cause or default), agency, vendor and dollars involved

Code: MIT Ā· Data: public domain (US government works)

DatasetWebsiteCodeHandoff guide

SF-133 Budget Execution

How much of each federal accountā€™s money has been spent, month by month.

  • Refreshes weekly
  • Live site
Size
3.06 million rows, FY2018 to FY2026

What you get

  • For each account: money available, not yet committed, committed and paid out
  • A flag for gift and donation accounts

Code: MIT Ā· Data: US government works

DatasetWebsiteCodeRebuild notes

USAJOBS Questionnaires

The questions on USAJOBS application questionnaires.

  • Refreshes daily
  • Live site
Size
52,227 questionnaires and 949,370 questions (2025)

What you get

  • Every questionnaire attached to a USAJOBS announcement since 2025, with the questions in order and their answer choices (email addresses removed)

Code: MIT Ā· Text: check the source terms

DatasetWebsiteCodeHandoff guide

šŸ‘‹ Next steps

Come say hi.

  • Poke around the repos and read the handoff guides.
  • Clone one and run it.
  • Play with the data on Hugging Face.
  • If you're interested, or just curious, email me at abigail.haddad@gmail.com. I'd love to hear from you.