RozgarMatch

The Problem

Fresh graduates in Pakistan face hundreds of generic listings and no signal about fit. They either spray applications everywhere or miss roles they were genuinely qualified for. Job boards rank by recency or ads, not by the distance between a graduate’s skills and a role’s requirements.

What I Built

• A FastAPI service that continuously aggregates live listings from public job feeds (RemoteOK and Arbeitnow), normalizes them, and filters to genuine tech roles.
• A transparent fit scorer: every result shows a fit percentage plus the matched and missing skills, so the ranking is explainable — never a black box.
• A clean web interface served by the same FastAPI app.

Technical Approach

• Ingestion: scheduled fetching with deduplication and title-based role filtering; in testing, ~400 fetched listings resolved to 164 relevant tech roles.
• Matching: skill extraction from job text matched against the graduate’s profile, weighted by requirement frequency.
• Design rule: a serious, genuinely needed product — real users, real listings, honest scores — not a dataset demo.

Technologies & Stack

Results

Live and public at rozgarmatch.onrender.com: paste your skills and it ranks current listings with a transparent fit score, showing which skills matched and which are missing. The full source is public on GitHub.

Limitations

Current listing sources skew remote and international; Pakistan-specific boards are the next integration. The fit score is a heuristic ranking aid, not a hiring guarantee.