JobAI Web: One Search Across Public Job Feeds

The Brief

Searching several public job boards separately makes it easy to lose track of where a listing came from, whether it is still open, and which link to use to apply. JobAI Web was built as a browser-based discovery tool: search by role and location in one place, then follow the original source link to apply. It does not host applications, guarantee vacancies, or manufacture listings.

How the Search Works

A responsive Flutter Web interface calls a Cloudflare Worker that gathers listings from public feeds, including Arbeitnow, Jobicy, Remote OK, Remotive, Himalayas, The Muse, and We Work Remotely. The Worker normalizes fields, deduplicates listings, and caches its collected feed for one hour. Greenhouse and Lever boards can be included when their public board slugs are configured. Each result retains its source attribution and an outbound link to the original job post.

Location Fallback, Not Profile Matching

JobAI Web first searches for the requested location. If that returns nothing, its optional AI-assisted step can expand the location to a nearby region; it can then fall back to relevant remote jobs worldwide. That step broadens a query rather than inventing vacancies or scoring a candidate against a profile. Results still depend on the actual public feeds and can change or expire.

Built for Everyday Browser Use

The interface supports desktop and mobile layouts, saved jobs, search history, dark mode, and PWA metadata. The public site is deployed on Cloudflare Pages at findjob.zaidanlab.dev; its Worker backend runs separately, and any optional model credential stays in the Worker environment rather than being compiled into browser code. The site publishes a readable introduction, its own sitemap, and correct 404 responses for unknown URLs.

The Outcome

JobAI Web is live and its search endpoint returns real, source-attributed listings. It demonstrates a small but practical architecture: a fast browser client, one API that handles multiple inconsistent feeds, and honest fallback behavior when a narrow search has no results.