create_sample_data.py

"""
create_sample_data.py
Generates 15 realistic sample resume PDFs for pipeline testing.
Profiles are intentionally varied so every pipeline stage gets exercised:
  - Some fail CSP (wrong degree, low exp, missing skills)
  - Some pass CSP but score low semantically
  - Some are strong matches (high composite + suitability score)

Run once: python create_sample_data.py
"""

import os
from fpdf import FPDF

os.makedirs("resumes", exist_ok=True)

# ── Candidate profiles ────────────────────────────────────────
# Format: (filename, full resume text)
CANDIDATES = [

    # ── STRONG MATCHES (should rank 1-5) ──────────────────────
    (
        "aisha_khan.pdf",
        """Aisha Khan
aisha.khan@email.com | Lahore, Pakistan

EDUCATION
Bachelor of Science in Computer Science - LUMS, 2021

EXPERIENCE
Machine Learning Engineer - DataFlow AI, 2021-Present (3 years of experience)
- Developed NLP pipelines using Python and spaCy for resume parsing
- Built classification models using scikit-learn and TensorFlow
- Deployed ML APIs using FastAPI and Docker on AWS

SKILLS
Python, machine learning, scikit-learn, TensorFlow, NLP, deep learning,
SQL, pandas, numpy, FastAPI, Docker, AWS, data analysis, REST API

PROJECTS
- Resume Screening Tool: TF-IDF + cosine similarity based candidate matcher
- Sentiment Analysis: BERT-based model fine-tuned on customer reviews
""",
    ),
    (
        "omar_farooq.pdf",
        """Omar Farooq
omar.farooq@email.com | Karachi, Pakistan

EDUCATION
Master of Science in Artificial Intelligence - NED University, 2020

EXPERIENCE
Senior ML Engineer - TechVentures, 2020-Present (4 years of experience)
- Led team of 3 to build production NLP systems
- Designed data preprocessing pipelines and feature engineering workflows
- Applied machine learning and deep learning models to real-world datasets

SKILLS
Python, machine learning, deep learning, PyTorch, scikit-learn, NLP,
computer vision, SQL, data analysis, pandas, numpy, Keras, OpenCV

CERTIFICATIONS
- DeepLearning.AI TensorFlow Developer Certificate
""",
    ),
    (
        "sara_malik.pdf",
        """Sara Malik
sara.malik@email.com | Islamabad, Pakistan

EDUCATION
Bachelor of Engineering in Software Engineering - NUST, 2022

EXPERIENCE
Junior ML Engineer - AI Solutions Pvt Ltd, 2022-Present (2 years of experience)
- Implemented machine learning models for predictive analytics
- Preprocessed large datasets using Python, pandas and NumPy
- Built REST APIs for model serving using Flask

SKILLS
Python, machine learning, scikit-learn, pandas, numpy, SQL, Flask,
data analysis, NLP, REST API, Git, Linux

PROJECTS
- Job Recommendation System using TF-IDF and cosine similarity
- Spam Classifier using Naive Bayes and logistic regression
""",
    ),
    (
        "bilal_ahmed.pdf",
        """Bilal Ahmed
bilal.ahmed@email.com | Rawalpindi, Pakistan

EDUCATION
Master of Science in Data Science - FAST NUCES, 2019

EXPERIENCE
Data Scientist - Insight Analytics, 2019-Present (5 years of experience)
- Designed end-to-end machine learning pipelines for client projects
- Applied NLP techniques including TF-IDF, word embeddings, and transformers
- Built and deployed deep learning models using TensorFlow and PyTorch

SKILLS
Python, machine learning, deep learning, TensorFlow, PyTorch, NLP,
scikit-learn, SQL, pandas, numpy, data analysis, Tableau, AWS, Spark

PUBLICATIONS
- "Comparative Analysis of Recruitment Matching Algorithms" - IEEE 2022
""",
    ),
    (
        "hina_javed.pdf",
        """Hina Javed
hina.javed@email.com | Lahore, Pakistan

EDUCATION
Bachelor of Science in Artificial Intelligence - COMSATS, 2023

EXPERIENCE
ML Research Assistant - VisionLab, 2023-Present (2 years of experience)
- Conducted research on computer vision and image classification
- Trained convolutional neural networks using PyTorch and Keras
- Preprocessed datasets using OpenCV and scikit-learn pipelines

SKILLS
Python, machine learning, deep learning, PyTorch, Keras, computer vision,
OpenCV, scikit-learn, numpy, pandas, NLP, SQL, Git

PROJECTS
- Object Detection System using YOLOv5
- Resume Parser using spaCy and regex
""",
    ),

    # ── MODERATE MATCHES (should rank 6-10) ───────────────────
    (
        "zain_raza.pdf",
        """Zain Raza
zain.raza@email.com | Multan, Pakistan

EDUCATION
Bachelor of Science in Computer Science - UET, 2021

EXPERIENCE
Software Developer - WebSoft Solutions, 2021-Present (3 years of experience)
- Developed web applications using React and Node.js
- Worked with SQL databases and REST APIs
- Some exposure to Python scripting and data processing

SKILLS
JavaScript, React, Node.js, SQL, REST API, Python, Git, Linux, HTML, CSS

PROJECTS
- E-commerce platform with real-time analytics dashboard
""",
    ),
    (
        "maryam_tariq.pdf",
        """Maryam Tariq
maryam.tariq@email.com | Faisalabad, Pakistan

EDUCATION
Bachelor of Science in Statistics - Punjab University, 2020

EXPERIENCE
Data Analyst - RetailMetrics Co, 2020-Present (4 years of experience)
- Performed data analysis and statistical modeling using R and Python
- Created dashboards using Power BI and Tableau
- Basic machine learning for sales forecasting

SKILLS
Python, R, SQL, data analysis, statistics, Power BI, Tableau,
pandas, numpy, Excel, machine learning

PROJECTS
- Sales Prediction Model using linear regression
- Customer Segmentation using K-means clustering
""",
    ),
    (
        "danish_ali.pdf",
        """Danish Ali
danish.ali@email.com | Hyderabad, Pakistan

EDUCATION
Bachelor of Engineering in Computer Systems - MUET, 2022

EXPERIENCE
Backend Developer - CloudApps, 2022-Present (2 years of experience)
- Built scalable APIs in Python using Django and FastAPI
- Managed PostgreSQL and MongoDB databases
- Deployed services on AWS using Docker and Kubernetes

SKILLS
Python, Django, FastAPI, SQL, PostgreSQL, MongoDB, Docker,
AWS, Kubernetes, REST API, Git, Linux

PROJECTS
- Microservices architecture for logistics platform
""",
    ),
    (
        "nadia_hussain.pdf",
        """Nadia Hussain
nadia.hussain@email.com | Peshawar, Pakistan

EDUCATION
Master of Science in Computer Science - UoP, 2021

EXPERIENCE
Research Associate - CogSystems Lab, 2021-Present (3 years of experience)
- Researched natural language processing and information retrieval
- Implemented text classification models using Python and scikit-learn
- Published 2 papers on NLP-based document matching systems

SKILLS
Python, NLP, scikit-learn, machine learning, text classification,
SQL, pandas, numpy, research, data analysis, spaCy

PROJECTS
- Document Similarity System using TF-IDF and BERT embeddings
""",
    ),
    (
        "hamza_sheikh.pdf",
        """Hamza Sheikh
hamza.sheikh@email.com | Quetta, Pakistan

EDUCATION
Bachelor of Science in Software Engineering - BUITEMS, 2023

EXPERIENCE
Junior Developer - AppWorks, 2023-Present (1 year of experience)
- Assisted in building ML features for a mobile application
- Python scripting for data processing automation
- Familiar with machine learning concepts and scikit-learn

SKILLS
Python, machine learning, scikit-learn, SQL, Git, pandas, Java, Android

PROJECTS
- Student Grade Prediction using decision tree
""",
    ),

    # ── CSP FAILURES (filtered out early) ────────────────────
    (
        "ali_rauf.pdf",
        """Ali Rauf
ali.rauf@email.com | Sialkot, Pakistan

EDUCATION
Diploma in Information Technology - Govt Polytechnic, 2020

EXPERIENCE
IT Support Technician - LocalTech, 2020-Present (4 years of experience)
- Managed hardware, networking, and desktop support
- Basic Python scripting for automation tasks

SKILLS
Python, networking, hardware, Linux, SQL, troubleshooting, Windows

NOTE: No formal bachelor degree completed yet.
""",
    ),
    (
        "farah_naz.pdf",
        """Farah Naz
farah.naz@email.com | Gujranwala, Pakistan

EDUCATION
Bachelor of Arts in English Literature - University of Punjab, 2021

EXPERIENCE
Content Writer - MediaGroup, 2021-Present (3 years of experience)
- Wrote technical blogs and documentation
- SEO optimization and content strategy

SKILLS
Writing, SEO, WordPress, Content Management, Microsoft Office, Research

NOTE: No technical programming skills or ML background.
""",
    ),
    (
        "tariq_mehmood.pdf",
        """Tariq Mehmood
tariq.mehmood@email.com | Sargodha, Pakistan

EDUCATION
Bachelor of Science in Computer Science - Virtual University, 2023

EXPERIENCE
Freelance Web Designer - Self-employed, 2023-Present (1 year of experience)
- Designed websites using HTML, CSS, and WordPress
- Basic JavaScript and PHP development

SKILLS
HTML, CSS, JavaScript, PHP, WordPress, Figma, Adobe XD

NOTE: Less than 2 years of experience, no machine learning background.
""",
    ),
    (
        "ayesha_siddiqui.pdf",
        """Ayesha Siddiqui
ayesha.siddiqui@email.com | Bahawalpur, Pakistan

EDUCATION
Bachelor of Science in Accounting - IBA Sukkur, 2020

EXPERIENCE
Accountant - FinancePro Pvt Ltd, 2020-Present (4 years of experience)
- Financial reporting, auditing, and bookkeeping
- Excel-based data analysis for finance team

SKILLS
Excel, accounting, financial analysis, bookkeeping, Tally, QuickBooks

NOTE: Finance background only, no ML or Python experience.
""",
    ),
    (
        "usman_butt.pdf",
        """Usman Butt
usman.butt@email.com | Lahore, Pakistan

EDUCATION
Master of Science in Electrical Engineering - UET Lahore, 2018

EXPERIENCE
Embedded Systems Engineer - AutoSys, 2018-Present (6 years of experience)
- Designed firmware for microcontrollers using C and C++
- Signal processing and hardware debugging
- Some Python scripting for test automation

SKILLS
C, C++, Python, MATLAB, embedded systems, signal processing,
microcontrollers, hardware debugging, PCB design

NOTE: Strong engineering background but no machine learning or NLP skills.
""",
    ),
]


def create_resume_pdf(filename: str, content: str, output_dir: str = "resumes"):
    """Creates a PDF resume from plain text content."""
    from fpdf import FPDF, XPos, YPos

    pdf = FPDF()
    pdf.add_page()
    pdf.set_margins(15, 15, 15)
    pdf.set_auto_page_break(auto=True, margin=15)
    W = pdf.w - pdf.l_margin - pdf.r_margin  # effective text width

    lines = content.strip().split("\n")

    # First line = name
    pdf.set_font("Helvetica", "B", 15)
    pdf.cell(W, 10, lines[0].strip(), new_x=XPos.LMARGIN, new_y=YPos.NEXT)

    # Second line = contact
    pdf.set_font("Helvetica", "", 9)
    if len(lines) > 1:
        pdf.cell(W, 6, lines[1].strip(), new_x=XPos.LMARGIN, new_y=YPos.NEXT)

    pdf.ln(3)

    for line in lines[2:]:
        line = line.strip()
        if not line:
            pdf.ln(3)
        elif line.isupper() or (line.endswith(":") and len(line) < 40):
            pdf.set_font("Helvetica", "B", 11)
            pdf.ln(2)
            pdf.cell(W, 7, line, new_x=XPos.LMARGIN, new_y=YPos.NEXT)
        elif line.startswith("-"):
            pdf.set_font("Helvetica", "", 10)
            pdf.multi_cell(W, 6, "  " + line)
        else:
            pdf.set_font("Helvetica", "", 10)
            pdf.multi_cell(W, 6, line)

    out_path = os.path.join(output_dir, filename)
    pdf.output(out_path)
    return out_path


if __name__ == "__main__":
    print("Generating sample resume PDFs...\n")
    for filename, content in CANDIDATES:
        path = create_resume_pdf(filename, content)
        print(f"  Created: {path}")

    print(f"\nDone! {len(CANDIDATES)} resumes generated in resumes/")
    print("\nProfile breakdown:")
    print("  Strong matches (expect rank 1-5) : aisha_khan, omar_farooq, sara_malik, bilal_ahmed, hina_javed")
    print("  Moderate matches (rank 6-10)     : zain_raza, maryam_tariq, danish_ali, nadia_hussain, hamza_sheikh")
    print("  CSP failures (filtered out)      : ali_rauf, farah_naz, tariq_mehmood, ayesha_siddiqui, usman_butt")