GPT-3 based resume parser as a REST API that transforms a resume PDF like this to a JSON like this.
Parsing a resume PDF takes around 15 seconds and costs about $0.01 for every 500 tokens using text-davinci-002
engine (that's why there is no live demo website). Note that a typical request and response may use 1500 tokens ($0.03), 3000 tokens ($0.06) or more.
Please note that more accurate results may be achieved by fine-tuning GPT-3, but the out-of-the-box results from this repo are already very impressive.
- Install Python 3 and pip3. For macOS, see note below.
- Install all dependencies of
pdftotext
(see here). - In a new terminal, update pip3 if needed:
python3 -m pip install --upgrade pip
- In another new terminal, clone the repository and move Terminal to the directory.
- Please close the other terminals and continue in this terminal.
- Check the versions:
python3 --version
andpip3 --version
. - Run the
./build.sh
in the project root. - Get your OpenAI API Key.
- Create a file named
.env
and set your API key in it:OPENAI_API_KEY=YOURKEY
or set the key in an environment variable:export OPENAI_API_KEY=YOURKEY
. - Run
./run.sh
in the project root.
A Flask server will start listening to port 5001 of localhost. Feel free to check it out with your browser.
You need to install either XCode or GCC tools (see here).
- If you install XCode, make sure to run it to complete the setup.
- Then run
xcode-select --install
and complete command-line tools installation. - Finally install Homebrew, and use
brew install python
to install Python 3.
- Basic Information
- first name
- last name
- full name
- U.S. phone number
- location
- portfolio website URL
- LinkedIn URL
- GitHub main page URL
- Education
- university
- education level
- graduation year
- graduation month
- majors
- GPA
- Job Experience
- job title
- company
- location
- duration
- job content
- Project Experience
- project name
- project description