job4u – AI Analysis of Job Postings
A traceable rationale instead of a black-box score
Hundreds of postings, no time: job4u reads every ad in full text and justifies its verdict instead of just scoring. 870 checked, 141 solid matches.
The problem
Anyone searching seriously faces hundreds of postings that nobody reads in full. Common job portals compare keywords and output a number — why a position fits or does not stays invisible. And the real work only starts afterwards: cover letter, CV, follow-up, keeping track.
- Hundreds of postings, no honest pre-selection
- Scores without a visible reason
- Application documents rebuilt by hand every time
- No overview of what went where and when
The solution
job4u reads every posting in full and checks it against a stored profile of evidenced facts. Every rating carries a rationale — you can read which requirement is met and which is not. From a match it builds a complete application package at the push of a button: cover letter, matching CV, notes on the recipient, stored with versioning.
My role
Sole developer and user
- Integration of the federal job search API and further sources
- Scoring logic with written rationale
- Profile and evidence management
- Generation of complete application packages
- Versioned storage per position
Features
Rationale instead of a score
Every rating names met and missing requirements in plain words.
Multiple sources
Federal job search, further portals and career pages in one pool.
Evidenced facts only
The profile separates evidenced from unevidenced. Missing experience is named, not claimed.
Application package at one click
Cover letter and CV tailored to the posting, ready to send as PDF.
Versioned storage
One folder per position with posting, package and notes — every version stays traceable.
Radius and reachability
Distance per position, checks whether a posting is still live, company profiles harvested from postings.
What I learned
The most important decision in job4u was a decision to leave something out: no score without a rationale. As soon as a system outputs a number, people believe it — even when nothing stands behind it. Force it to explain every rating and the weak judgements expose themselves. The same principle runs through my other systems: verify first, claim second.