CareerBert is a tool that brings a data-driven approach to your job search. The aim of this project is to develop a different approach for comparing multiple job descriptions with a given resume.
This system uses the power of DistilBertSequence Classification to classify sentences into categories such as job description, skills, and education. The model training data labeling was accomplished using Doccano, an open-source text annotation tool, to label sentences effectively.
data science, data analysis, and data engineering roles.
For data extraction, an API key from SerpAPI is required. Once obtained, replace the api_key on line 46 in the web_scrape.py file. This will result in a CSV file containing data for 50 jobs.
❗ Run judiciously to avoid ban due to scraping rules.
Model training requires GPU support and is done on Google Colab with a GPU runtime. After data scraping, load the data into the Google Colab notebook to train the model.
To train a model on a different role, uncomment the last line in the notebook that saves a model, add the model to the trained_models directory, and change the model name in the code accordingly.
The system transforms each sentence of a resume and the job description in question into a TF-IDF representation and uses cosine similarity to compare it with job descriptions. Resume parsing is currently supported only for PDF files.
Before running the programs make sure to run in terminal: pip install -r requirements.txt
Choose between running in the terminal or launching a Flask app:
- For terminal:
python main.py resume.pdf - For Flask app:
python app.py
The Flask app then displays the job description, the link to the webpage, and the similarity scores in total and per section.
Currently, the resume parser expects the resume sections to be in the following format:
- EDUCATION
- EXPERIENCE
- PERSONAL PROJECTS
- PUBLICATIONS
- TECHNICAL SKILLS
This will expand in the future to accommodate more diverse resume structures.