quantyx
The Indian data & AI job market, re‑derived every morning.
Four job APIs in, raw data committed to git, dbt on DuckDB out — republished only when it passes its own tests. Nobody touches it.
- assert_volume_not_collapsed Today’s volume against its trailing seven-day average Armed
- assert_no_source_disappeared A channel that goes silent while the others carry the total Armed
- assert_required_fields_present A field that stops being populated after an upstream rename Armed
- assert_compensation_is_coherent Pay figures that cannot be true Armed
All four passed on 2026-09-11, which is the only reason this page updated.
Any one of them returning a row stops the deploy, files an issue, and leaves the previous dashboard live — so a stale page, not a wrong one, is what failure looks like here.
scripts/verify_drift_detection.sh corrupts the sample dataset on purpose in CI
and asserts these tests fail. A test never observed failing is not evidence of anything.
Market activity
40 day(s) recordedRoles open on each day the pipeline has run, and the roles it had never seen before. New listings are the cleaner signal: the open total moves slowly, discoveries spike.
Remote against on-site
What employers ask for
Top 12 of 109Every skill is matched by a hand-written regular expression against the posting's title and description, so any number here traces back to a pattern you can read. There is no model guessing in the loop.
How to check this
The repository is the dataset. Every posting the pipeline has ever seen is committed as gzipped JSONL, so every number above can be recomputed from scratch by anyone who clones it. Nothing here is scraped from a site whose robots.txt disallows it — an earlier plan to scrape Internshala was abandoned when its robots.txt turned out to forbid exactly the pages it needed.
dbt build Rebuilds every figure on this site from the committed raw data.
scripts/verify_drift_detection.sh Corrupts the sample dataset on purpose and asserts the drift tests fail.
git log The daily commits are the pipeline's own. It writes its data and its statistics without a human.