scholarweave/Jobs-Dataset-v2

OpenJobData Jobs Dataset

OpenJobData is an open dataset of job listings collected from company career pages. It includes a current view of open jobs, daily changes to listings, and detected closures so that you can explore hiring activity over time.

Use the minimal files for search, analysis, and most applications. Use the full files when you need the additional raw and normalized source data for auditing or reprocessing.

Links

• Browse the data
• Read the documentation
• Explore job-market statistics

Dataset layout

data/
├── manifests/
│   └── current.json
├── companies/
│   └── snapshot_date=YYYY-MM-DD/part-*.parquet
├── active/
│   ├── minimal/current/shard_0000.parquet
│   └── full/current/shard_0000.parquet
├── changes/
│   ├── minimal/date=YYYY-MM-DD/run_id=<run-id>/part-*.parquet
│   └── full/date=YYYY-MM-DD/run_id=<run-id>/part-*.parquet
└── closed/
    ├── minimal/date=YYYY-MM-DD/run_id=<run-id>/part-*.parquet
    └── full/date=YYYY-MM-DD/run_id=<run-id>/part-*.parquet

Folder	What it contains
active/	Current open job listings. Read the files named in manifests/current.json.
changes/	Jobs that were added or materially updated on a given day.
closed/	Jobs that were detected as no longer open.
companies/	Company lookup snapshots, when available.
manifests/current.json	The list of active files for a consistent current snapshot.

Getting started

Install the tools you need:

pip install huggingface_hub duckdb pyarrow

Query current jobs with DuckDB

DuckDB can query the public Parquet files directly:

INSTALL httpfs;
LOAD httpfs;

SELECT title, company_id, country, apply_url
FROM 'hf://datasets/scholarweave/Jobs-Dataset-v2/data/active/minimal/current/*.parquet'
WHERE is_remote = true
LIMIT 100;

Download the current-job manifest

Read current.json first when you need a reproducible, integrity-checked snapshot. It identifies the active Parquet shards that belong together.

import json
from huggingface_hub import HfFileSystem

repo = "scholarweave/Jobs-Dataset-v2"
fs = HfFileSystem()

with fs.open(f"datasets/{repo}/data/manifests/current.json", "rb") as file:
    manifest = json.load(file)

for shard, entry in manifest["active"]["minimal"]["shards"].items():
    print(shard, entry["path"])

Main fields

Fields vary by source and dataset variant. Current minimal files commonly include:

• posting_id — the source job identifier
• lifecycle_id and lifecycle_sequence — identifiers for tracking a job over time
• company_id, title, department, and employment_type
• workplace_type, country, and is_remote
• posted_at, apply_url, first_seen_at, and last_seen_at
• description_plain, salary, location, and skills fields when supplied by the source

The full variant retains additional JSON fields for auditing and reprocessing. All timestamps are UTC. A source-provided posted_at value may be incomplete or inaccurate; first_seen_at and last_seen_at indicate when OpenJobData observed a listing. A detected closed_at is not necessarily the time an employer closed the role.

Recommended uses

• Job-search tools and job boards
• Labor-market and hiring-trend research
• Company and role discovery
• Historical analysis of listing changes and closures
• Reproducible research using a pinned Hugging Face revision

Limitations and responsible use

Job sources can change, remove, duplicate, or misclassify listings. Check the schema of the files you use, especially for optional fields that vary by applicant tracking system. For reproducible work, pin a Hugging Face commit SHA and record the schema_version used in your analysis.

Respect the dataset license, source terms, applicable laws, and privacy obligations when using this dataset.