Job Finder

A FastAPI + React web app that turns a plain-language brief, “video game studios in Montpellier with fewer than 50 employees”, into a structured, evidence-backed list of French companies. A Web Information Retrieval project at Tsinghua University.

A search runs as a transparent, resumable pipeline.

From a sentence to qualified companies

The brief is parsed into structured filters and run through a four-step research console:

1. Map the request to NAF 2025 codes, then crosswalk them to the older rev.2 activity codes used by the public company API.
2. Retrieve matching establishments from the French company registry.
3. Score relevance and surface candidates for manual review.
4. Enrich each kept company with web evidence: contacts, public signals and the sources behind them.

Writing the natural-language brief.

Reviewing and tuning the inferred filters before retrieval.

Scored candidates laid out for manual review.

Evidence and exploration

Results come with the evidence that justifies them, a map overview, and per-company detail.

Geographic overview of the matched companies.

Per-company detail with contacts and public signals.

Enriched contact information with sources.

Built to resume and stream

Searches are persisted in MongoDB, so they are resumable, and the UI streams live progress through Server-Sent Events. It runs as both a web app and a CLI.

My role

Full-stack developer and WebIR pipeline engineer: the retrieval/scoring/enrichment pipeline, the FastAPI backend and the React front end.

Tech

Python · FastAPI · React · MongoDB · Server-Sent Events