Daily snapshots of data science / machine learning job postings scraped from LinkedIn, one parquet file per scrape run. All splits share one schema (2023 → today); historical splits were migrated in 2026-07 — the original 8-column data is preserved at revision v1-schema.
The same job_id recurs across splits (a posting stays live for days) — that's the longitudinal signal. For a unique-jobs view, dedup on job_id keeping the row with max scrape_dt.… See the full description on the dataset page:
https://huggingface.co/datasets/ryang2/linkedin-job-scrape.
