Jobs Engine · 2026
Rank public listings against a profile
Laboratory
- Role
- Independent engineer
- Stack
- Python, FastAPI
- Status
- Laboratory
- Live
- scrocle.cloud/demo/jobs-engine
Problem
A hiring workflow starts with a profile and a pile of listings. Someone has to score them, drop the misses, and put the rest in a sheet. Doing that by hand is slow. Doing it with a black-box model is hard to explain.
Constraint
This demo does not scrape LinkedIn, Indeed, or Upwork. It does not call a language model. It does not write to Google Sheets. Listings are a captured sample from Himalayas, Jobicy, and Remotive public feeds.
Architecture
- Title / skills / location
- Keyword rubric
- Ranked cards
- Himalayas / Jobicy / Remotive sample
- Excel download
Decisions
Show the score parts
Skills, title, and location are separate bars. If FastAPI is missing, you can see it. A single magic number would hide that.
Fixture, not a live scrape
Public APIs fail and mix in junk. A frozen set means the page always has rows. Production would refresh the same feeds on a schedule.
Open a preset. The list re-ranks. Download the sheet if you want the same rows in Excel.
If you already have boards you care about and a Sheet the team lives in, that is the production slice. This page is the ranking and export part, on public feeds, with the model turned off so you can see the score.
Outcome
Demo. Live at scrocle.cloud/demo/jobs-engine. Not a client job.
What was handed over
- How to run the FastAPI demo locally
- That matching is keyword overlap, not a model
- What a production refresh and Google Sheet would add