AI Job Search Agent
by: Tomás Ospina.
Overview
An automated job-search agent that reads a CV, builds a structured professional profile from it, pulls openings from multiple platforms, and scores each one from 0 to 100 — together with why the candidate fits, where the competitive advantage is, and what the concerns are.
Problem & Solution
Job boards rank by keyword and recency, which puts the interesting 5% of postings somewhere in the middle of page four. This agent inverts the process: the profile is extracted once from the CV, and every posting is then scored against it, so the ranking reflects fit rather than search relevance.
It handles two distinct profiles — AI Automation / Backend and Embedded Systems / Hardware — and scores each posting against the relevant one, since a firmware role and a backend role reward completely different parts of the same CV.
Technical Specifications
| Component | Detail |
| —————- | ——————————————————— |
| cv_parser.py | Extracts a structured profile from the CV PDFs |
| job_sources.py / job_search_fetcher.py | Multi-platform posting retrieval |
| job_search_agent.py | Matching and scoring, 0–100 with rationale |
| database.py | SQLite (jobs.db) — postings, scores and history |
| scheduler.py | Periodic re-runs so new postings are scored automatically |
| app.py | Flask interface with the ranked results |
My Role & Key Contributions
- Built the whole pipeline, from PDF parsing to the web interface.
- Designed the scoring model so every number comes with an explanation — strengths, competitive advantages and concerns — rather than an opaque percentage.
- Added the scheduler so the database stays current without manual runs.
What I Learned
A score is only useful if it is auditable. Making the agent explain each ranking was what turned it from a toy into something worth acting on.
🔒 Private repository — happy to walk through the code on request.