DEET Automatch

Upload a resume. Get a complete DEET profile in seconds — plus a live board of verified job vacancies scraped directly from employer career pages.

Built for the DEET Hackathon · February 2026

---

What it does

Most job portals make applicants re-type everything from their resume. This fixes that.

• Resume → Profile: Upload a PDF, DOCX, or image. The app extracts your name, contact details, education, skills, work experience, and projects using OCR + NLP, then shows you a confidence-scored editable preview before saving.
• Job Discovery: A background scheduler scrapes verified employer career pages every 6 hours, deduplicates listings, classifies them by category, and surfaces them on a filterable job board.

---

Demo

---

Tech stack

Layer	Tools
Backend	Flask, SQLite, APScheduler
OCR	pdfplumber (native PDF), pytesseract (scanned/image)
NLP	spaCy (NER), regex, python-dateutil
ML	scikit-learn — TF-IDF + Logistic Regression
Scraping	BeautifulSoup4, requests
Verification	python-whois, DNS resolution
Testing	pytest

---

Getting started

Requirements: Python 3.10+ — Download here (check "Add to PATH" during install)

Windows

# First-time setup
setup.bat

# Start the server
run.bat

Mac / Linux

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python -m spacy download en_core_web_sm
python app.py

Then open http://localhost:5000 in your browser.

---

How the pipelines work

Resume → DEET profile

Upload (PDF / DOCX / Image)
        │
        ▼
   OCR Engine
   ├── Native PDF  → pdfplumber
   ├── Scanned PDF → pytesseract
   └── Image       → PIL + pytesseract
        │
        ▼
   NLP Extractor
   ├── Section detection  (regex headers)
   ├── Named Entity Recognition  (spaCy)
   ├── Pattern matching  (email, phone, LinkedIn, GitHub)
   ├── Skills extraction  (keyword dictionary)
   └── Date normalization  (dateutil)
        │
        ▼
   Confidence Scorer  →  HIGH / MEDIUM / LOW badge per field
        │
        ▼
   Editable preview  →  User confirms  →  Saved to DB

Extracted fields: Name · Email · Phone · Address · LinkedIn · GitHub · Summary · Education · Skills · Certifications · Experience · Projects

---

Job vacancy discovery

APScheduler (every 6h) or manual trigger
        │
        ▼
   Scraper  (BeautifulSoup per career page)
        │
        ▼
   Deduplicator
   ├── Exact  →  SHA-256 hash of title + company + location
   └── Near-duplicate  →  TF-IDF cosine similarity ≥ 0.85
        │
        ▼
   Classifier  →  TF-IDF + Logistic Regression  (keyword fallback)
        │
        ▼
   Employer Verifier
   ├── DNS resolution   +0.25
   ├── HTTP reachable   +0.25
   ├── HTTPS            +0.20
   └── Domain age ≥ 1yr +0.30
        │
        ▼
   SQLite  →  Job board

---

API reference

Method	Endpoint	Description
GET	/	Landing page
GET	/jobs	Job discovery board
GET	/preview	DEET profile editor
POST	/api/resume/upload	Upload resume → extracted profile JSON
POST	/api/resume/submit	Save confirmed profile
GET	/api/resume/profiles	List all saved profiles
GET	/api/jobs/list	Paginated jobs (filter by category, location, search)
POST	/api/jobs/crawl	Trigger immediate crawl
POST	/api/jobs/add	Manually add a job posting
GET	/api/dashboard/stats	Platform statistics
GET	/api/health	Health check

---

Configuration

Edit config.py:

Setting	Default	Description
CRAWL_INTERVAL_HOURS	6	Auto-crawl frequency
DEDUP_COSINE_THRESHOLD	0.85	Near-duplicate sensitivity
MIN_DOMAIN_AGE_DAYS	180	Employer trust threshold
TARGET_CAREER_PAGES	[]	Career page URLs to scrape

Adding a career page

# config.py
TARGET_CAREER_PAGES = [
    {
        "company":           "Acme Corp",
        "url":               "https://acme.com/careers",
        "job_selector":      ".job-listing",
        "title_selector":    ".job-title",
        "location_selector": ".job-location",
        "link_selector":     "a",
    },
]

---

Running tests

# Run all tests
python -m pytest tests/ -v

# Run a specific module
python -m pytest tests/test_nlp_extractor.py -v
python -m pytest tests/test_api.py -v

---

Project structure

├── app.py                     # Flask entry point
├── config.py                  # Configuration
├── database.py                # SQLite helpers
├── requirements.txt
├── modules/
│   ├── ocr_engine.py          # PDF + image text extraction
│   ├── nlp_extractor.py       # NLP pipeline
│   ├── confidence_scorer.py   # Per-field confidence (0.0–1.0)
│   ├── job_scraper.py         # Career page scraper
│   ├── job_classifier.py      # ML job category classifier
│   ├── deduplicator.py        # Duplicate detection
│   ├── employer_verifier.py   # Domain trust scoring
│   └── scheduler.py           # Background crawl scheduler
├── templates/
│   ├── index.html             # Upload UI
│   ├── preview.html           # Profile editor
│   └── jobs.html              # Job board
├── static/
│   ├── css/style.css
│   └── js/
│       ├── main.js
│       ├── preview.js
│       └── jobs.js
└── tests/
    ├── test_nlp_extractor.py
    ├── test_confidence_scorer.py
    ├── test_deduplicator.py
    └── test_api.py

---

Target metrics

Metric	Target
Field extraction F1	≥ 85%
OCR character error rate	≤ 5%
Job discovery precision	≥ 90%
Duplicate detection rate	≥ 95%
Resume API latency (p95)	≤ 5s

---

License

MIT