💼 Job Match Intelligence System

An end-to-end intelligent job matching platform that collects real-world job postings, matches them against candidate profiles using an explainable scoring engine, and serves everything through a live web application.

🔗 Live App: job-match-intelligence-system-m7srotvkvxfkmwhm9e35kv.streamlit.app

---

🚀 What It Does

• Collects live job postings via the JSearch API (LinkedIn, Indeed, Glassdoor, and more)
• Extracts job requirements — skills, experience, education, seniority
• Parses candidate profiles (manual input or resume upload — PDF/DOCX)
• Computes explainable match scores with component breakdowns
• Generates personalized job recommendations ranked by fit
• Full authentication system — register, login, JWT, password reset via email
• Multi-profile support — manage multiple candidate profiles per account
• Google Analytics 4 tracking with UTM attribution

---

🏗️ Architecture

Raw Jobs → Staging → Curated → Extracted → Matching Engine → FastAPI → Streamlit UI

Backend: FastAPI + SQLAlchemy + SQLite — deployed on Render
Frontend: Streamlit — deployed on Streamlit Cloud

---

📁 Project Structure

Job-Match-Intelligence-System/
│
├── app/
│   └── streamlit_app.py        # Full Streamlit frontend
│
├── configs/
│   ├── sources.yaml            # API keys & email config
│   ├── skills.yaml             # Skill taxonomy
│   └── scoring.yaml            # Scoring weights
│
├── src/
│   ├── api/
│   │   ├── main.py             # FastAPI app + all endpoints
│   │   └── email_service.py    # SMTP password reset emails
│   ├── auth/                   # JWT authentication
│   ├── candidate/
│   │   └── resume_extractor.py # PDF/DOCX resume parser
│   ├── db/                     # SQLAlchemy models
│   ├── extraction/             # Job requirement extraction
│   ├── ingestion/
│   │   └── jsearch.py          # Live JSearch API client
│   ├── matching/
│   │   ├── scoring.py          # Weighted match scoring engine
│   │   ├── ranker.py           # Job ranking
│   │   └── job_templates.py    # 150+ job title skill map
│   └── normalization/          # Data cleaning & structuring
│
├── requirements.txt            # Frontend deps (Streamlit Cloud)
├── requirements-backend.txt    # Backend deps (Render)
├── render.yaml                 # Render deployment config
├── .python-version             # Python 3.11.9
└── DEPLOY.md                   # Full deployment guide

---

⚙️ System Components

✔ Phase 1–4 — Data Pipeline

• Config-driven YAML architecture
• Job ingestion from Greenhouse & Lever (~747 real postings)
• Title, location, and text normalization
• Deduplication via hashing
• Requirement extraction: skills, experience, education, seniority

✔ Phase 5 — Candidate Understanding

• Structured candidate profile schema
• Skill, tool, domain, education normalization
• Seniority inference & keyword aggregation
• Resume parsing from PDF and DOCX files

✔ Phase 6 — Matching Engine

Hard Filters: required skills · experience · education

Weighted Scoring:

Component	Weight
Required Skills	High
Preferred Skills	Medium
Experience	Medium
Education	Low
Seniority	Medium
Domain Alignment	Medium

Output per job:

• Match score (0–100) + Fit label (Strong / Good / Partial / Weak)
• Matched skills, missing skills, component breakdown

✔ Phase 7 — API Layer (FastAPI)

Category	Endpoints
Auth	POST /register · POST /login · POST /password-reset
Profiles	GET/POST /profiles · PUT/DELETE /profiles/{id}
Matching	POST /match · POST /recommendations
Live Jobs	POST /recommendations/live (JSearch)
Resume	POST /resume/parse

Swagger docs: https://job-match-api-iibv.onrender.com/docs

✔ Phase 8 — Frontend (Streamlit)

• Dark theme UI with colorful job cards
• Login / Register / Forgot Password / Change Password
• Candidate profile builder + resume upload
• Profile completeness progress bar
• Job Matches tab — dataset-based recommendations
• Live Jobs tab — real-time JSearch results with match scoring
• Score rings, skill chips (matched/missing), apply buttons

✔ Phase 9 — Evaluation

Metric	Score
Extraction Precision	0.75
Extraction Recall	0.625
Extraction F1	0.679
Matching Accuracy	66.7%

---

🖥️ Local Development

# Terminal 1 — Backend
uvicorn src.api.main:app --reload

# Terminal 2 — Frontend
streamlit run app/streamlit_app.py

Open http://localhost:8501 in your browser.

See DEPLOY.md for full deployment instructions.

---

💡 Key Strengths

• ✔ End-to-end system — data pipeline → matching engine → API → live web app
• ✔ Explainable AI — transparent, component-level scoring
• ✔ Real-world job data — live JSearch integration (LinkedIn, Indeed, Glassdoor)
• ✔ Resume parsing — upload PDF or DOCX to auto-fill profile
• ✔ Production deployed — Render (backend) + Streamlit Cloud (frontend)
• ✔ Analytics — GA4 + UTM tracking for visitor attribution

---

👨‍💻 Author

Ahmad Issa
Machine Learning Engineer | Data Science & AI Systems
GitHub · LinkedIn