SkillBridgeEngine: Contextual Skill Gap Analyzer

SkillBridgeEngine is a local, privacy-first machine learning pipeline designed to analyze a candidate's resume against a target job description. It extracts technical entities, computes the semantic "skill gap" using state-of-the-art vector embeddings, and dynamically generates a personalized syllabus of free courses to bridge that gap.

🧠 Core Architecture

• Data Ingestion: pdfplumber for robust PDF resume parsing without breaking multi-column layouts.
• Custom NER Pipeline: A locally trained spaCy model (built on en_core_web_md) to accurately extract modern tech entities (HARD_SKILL, SOFT_SKILL, TOOL).
• Semantic Engine: sentence-transformers utilizing BAAI/bge-base-en-v1.5 for high-accuracy, low-latency cosine similarity matching (e.g., matching "AWS" to "Amazon Web Services").
• Vector Database: Local ChromaDB for storing and retrieving high-dimensional course embeddings.

📂 Project Structure

├── mooc_dataset.csv          # Source data for course recommendations

⚙️ Setup & Installation

1. Install Dependencies Ensure you have Python 3.9+ installed, then run:

pip install pandas chromadb sentence-transformers spacy pdfplumber scikit-learn datasets

2. Download the Base NLP Model Download the medium English vocabulary model for spaCy:

python -m spacy download en_core_web_md

3. Prepare Dummy Data Place the following files in your root directory to test the pipeline:

• sample_resume.pdf (Your target resume)
• sample_jd.txt (The target job description in plain text)
• mooc_dataset.csv (Your library of available courses)

🚀 Usage

This project is divided into two phases: a one-time setup phase (training the models and populating the database) and an active inference phase (analyzing resumes).

Phase 1: System Setup

Run this command once to download the Hugging Face job descriptions, train the custom Named Entity Recognition (NER) model, and build the local ChromaDB vector space.

python main.py --setup

Note: This will create two new local folders: /custom_ner_model and /chroma_db.

Phase 2: Skill Gap Analysis

Once setup is complete, run the analysis engine against your resume and job description:

python main.py --analyze --resume sample_resume.pdf --jd sample_jd.txt

Expected Output: The terminal will output the extracted skills from both documents, isolate the missing skills using semantic cosine similarity, and print a top-3 recommended course syllabus fetched directly from your local vector database.

📝 Google Colab / Jupyter Note

If you are running this entirely inside a Jupyter Notebook or Google Colab environment, standard argparse CLI commands may conflict with the kernel. Bypass main.py and run the functions directly in your cells:

from main import run_setup, run_analysis

# 1. Run once
run_setup()

# 2. Run for each resume
run_analysis(resume_path="sample_resume.pdf", jd_path="sample_jd.txt")

---