Skip to content

jozsefszalma/interview_coach_GPT-4

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Job Interview Coach

Mock Interview with Feedback

image

Setup

Clone repo into a folder of your choice.

With Docker

run in the same folder:

  • docker build -t interview_bot .
  • docker run -p 7860:7860 --env KEY=your_openai_api_key interview_bot
  • navigate to http://127.0.0.1:7860/ with your browser

From Notebook

  • create a Python 3.9 venv or conda environment
  • install dependencies e.g. pip install -r requirements.txt
  • set up your OpenAI API key as "KEY" environment variable (e.g. via .env file if using VSCode)
  • As the UI might not fully render in a small window within an IDE I recommend connecting to the URL returned by Gradio, e.g. the default http://127.0.0.1:7860/ with your browser

Usage

provide your CV pdf on the left-most tab, provide the Job Description on the middle tab and begin the interview on the right-most tab.

LICENSE AND DISCLAIMER:

Copyright 2023, Jozsef Szalma
Creative Commons Attribution-NonCommercial 4.0 International Public License
Gradio code was partially reused from / informed by this guide

Before repurposing this code for an HR use-case consider:

  • OpenAI's useage policies expliclitly prohibit:
    "Activity that has high risk of economic harm, including [...] Automated determinations of eligibility for [...] employment [...]"

  • The EU AI Act proposal contains the following language:
    "AI systems used in employment, workers management and access to self-employment, notably for the recruitment and selection of persons [..] should also be classified as high-risk"

KNOWN ISSUES:

  • incomplete error handling around job description, e.g. if an invalid JD URL is provided the code won't fall back to the copy-pasted JD
  • if no JD and/or CV are provided GPT-4 might on occasion ignore instructions to only ask one interview question at a time
  • the current workflow consumes a lot of tokens as the JD and the CV aren't summarized, but considered as-is for each question
  • the scraping logic breaks once the job is in the "no longer accepting applications" status