ztrimus / job-llm

Simplify and improve the job hunting experience by integrating LLMs to automate tasks such as resume and cover letter generation, as well as application submission, saving users time and effort.

ResumeFlow: An LLM-facilitated Pipeline for Personalized Resume Generation and Refinement

For Video Demonstration visit the YouTube link: https://youtu.be/Agl7ugyu1N4

Project can be:

• Access as a Web Tool from https://job-aligned-resume.streamlit.app/
• Install as a Python Package from https://pypi.org/project/zlm/
• download as Source Code from https://github.com/Ztrimus/job-llm.git

All other known bugs, fixes, feedbacks, and feature requests can be reported on the GitHub issues page.

Empower others, just like they helped you! Contribute to this open source project & make a difference. ✨ Create a branch, improve the code, & raise a pull request!

Author & Contributor List

• Saurabh Zinjad | Ztrimus | [email protected]
• Amey Bhilegaonkar | ameygoes | [email protected]
• Amrita Bhattacharjee | Amritabh | [email protected]

1. Introduction:

1.1. Motivation: LLMs as Components in an ML Pipeline

In this project, we will investigate how to effectively use Large Language Models (LLMs) to automate various aspects of this pipeline.

Because, Solving a task using machine learning methods requires a series of steps that often require large amounts of human effort or labor. Furthermore there might be more steps after the training the ML model, such as evaluation, explaining the behavior of the model, interpreting model outputs, etc. Many of these steps are also often human labor intensive.

1.2. Our Proposal

We're aiming to create a automated system that makes applying for jobs a breeze. Job hunting has many stages, and we see a chance to automate things and use LLM (Language Model) to make it even smoother. We're looking at different ways, both the usual and some new ideas, to integrate LLM into the job application process. The goal is to reduce how much you have to do and let LLM do its thing, making the whole process easier for you.

1.3. Refer to this Paper for more details.

2. Setup, Installation and Usage

2.1. Prerequisites

• OS : Linux, Mac
• Python : 3.11.6 and above
• LLM API key: OpenAI OR Gemini Pro

2.2. Package Installation - Use as Library

pip install zlm

• Usage

from zlm import AutoApplyModel

job_llm = AutoApplyModel(
    api_key="PROVIDE_API_KEY", 
    provider="ENTER PROVIDER <gemini> or <openai>",
    downloads_dir="[optional] ENTER FOLDER PATH WHERE FILE GET DOWNLOADED, By default, 'downloads' folder"
)

job_llm.resume_cv_pipeline(
    "ENTER_JOB_URL", 
    "YOUR_MASTER_RESUME_DATA" # .pdf or .json
) # Return and downloads curated resume and cover letter.

2.4. Setup & Run Code - Use as Project

git clone https://github.com/Ztrimus/job-llm.git
cd job-llm

1. Create and activate python environment (use python -m venv .env or conda or etc.) to avoid any package dependency conflict.
2. Install Poetry package (dependency management and packaging tool)

   pip install poetry

3. Install all required packages.
   • Refer pyproject.toml or poetry.lock for list of packages.

     poetry install

     OR
   • If above command not working, we also provided requirements.txt file. But, we recommend using poetry.

     pip install -r resources/requirements.txt

4. We also need to install following packages to conversion of latex to pdf
   • For linux

     sudo apt-get install texlive-latex-base texlive-fonts-recommended texlive-fonts-extra

     NOTE: try sudo apt-get update if terminal unable to locate package.
   • For Mac

     brew install basictex
     sudo tlmgr install enumitem fontawesome

5. Run following script to get result

>>> python main.py /
    --url "JOB_POSTING_URL" /
    --master_data="JSON_USER_MASTER_DATA" /
    --api_key="YOUR_LLM_PROVIDER_API_KEY" / # put api_key considering provider
    --downloads_dir="DOWNLOAD_LOCATION_FOR_RESUME_CV" /
    --provider="openai" # openai, gemini, together, g4f

3. Citations

If you find JobLLM useful in your research or applications, please consider giving us a star 🌟 and citing it.

@misc{zinjad2024resumeflow,
      title={ResumeFlow: An LLM-facilitated Pipeline for Personalized Resume Generation and Refinement}, 
      author={Saurabh Bhausaheb Zinjad and Amrita Bhattacharjee and Amey Bhilegaonkar and Huan Liu},
      year={2024},
      eprint={2402.06221},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

4. License

JobLLM is under the MIT License and is supported for commercial usage.

4. References

• Prompt engineering Guidelines
• Overleaf LaTex Resume Template
• Combining LaTeX with Python
• OpenAI Documentation

job-llm's Issues

Vector Data Storage and Retrieval Workflow

Tasks to perform:

• Vector Data Storage
  • Ingestion of data or documents
  • Langchain installation and setup
  • Load data
  • Split the data
  • Create the embeddings
  • Store the chunks in vector store
• Information Retrieval
  • Create embedding of query/prompt
  • Function to retrieve similar embedding from vector store
• Integration this with existing pipeline

Overview of workflow

[Feature] Enable users to edit resume layouts, and content in the web app

Resume Customization Options: Easy to Advanced:

1. Choose sections to include: We can start by offering a multi-select dropdown field in the web app, allowing users to choose which sections of their resume they want to include.
2. Fine-tune details with Overleaf: we can give users the option to edit the LaTeX file generated by the job-llm tool. this can be done by directing the latex file to a third-party application like Overleaf. With Overleaf, users can modify the LaTeX file and compile it into a PDF format.
3. Pick and choose content: To provide more control, we can include checkbox options for users in the intermediate resume JSON file. Each checkbox corresponds to a key-value pair generated by the job-llm tool, allowing users to selectively include or exclude specific points in their resume.

Package project to ship as Python library

• Do setup change.
• Build code
• Publish as wheel file

Getting Unbound error

UnboundLocalError: This app has encountered an error. The original error message is redacted to prevent data leaks. Full error details have been recorded in the logs (if you're on Streamlit Cloud, click on 'Manage app' in the lower right of your app).
Traceback:
File "/home/adminuser/venv/lib/python3.11/site-packages/streamlit/runtime/scriptrunner/script_runner.py", line 535, in _run_script
exec(code, module._dict_)
File "/mount/src/job-llm/web_app.py", line 120, in <module>
resume_path, resume_details = resume_llm.resume_builder(job_details, user_data, is_st=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/mount/src/job-llm/zlm/utils/utils.py", line 100, in wrapper
result = func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/mount/src/job-llm/zlm/_init_.py", line 332, in resume_builder
return resume_path, resume_details
^^^^^^^^^^^

Create User Interface

• Install streamlit
• Give option to upload User data
• Give option to upload job details
• Button to download Resume
• Button to download Cover letter

Section Wise Resume Generation and LLM Prompting

• Instead of instructing LLM to generate the entire resume in one go, we will generate each section separately.
• This approach provides leverage in terms of context size and allows for specific section instructions during generation.

TO DO Tasks:

• Craft section-wise prompts
• Enhance resume_builder function

Option to Select different LLM Provider - Together.ai

• Get API keys
• Write class for integration of Together.ai functionality and api keys

Create REST API Endpoints

dependency-resolution: Poetry Dependency Resolution Issue with Python 3.11.7

When attempting to use the poetry install command with Python version 3.11.7, the following error is encountered:

The currently activated Python version 3.11.7 is not supported by the project (3.11.6).
Trying to find and use a compatible version.
Python was not found; run without arguments to install from the Microsoft Store, or disable this shortcut from Settings > Manage App Execution Aliases.

Poetry was unable to find a compatible version. If you have one, you can explicitly use it via the "env use" command.

Steps to Reproduce:

1. Activate Python version 3.11.7.
2. Run poetry install command

Implement Evaluation Metrics

• Explore different evaluation metrics for current use-case.
• Implement Jaccard similarity.

Stage 1 - Extracting Job Description Summary from URL

• Set Up Project Environment
  • Install Poetry
  • Install Required Python Packages
• Update README file
• Define Optimal System Prompt/Custom Instruction for AI Persona - Job Details Extractor
• Implement Data Extraction (Web Scraping) Functionality for Any URL
• Acquire OpenAI API Key
• Code for ChatGPT Integration
• Code for LLama2 Integration
• Develop Demo Script to Showcase Job Details Extraction Results

Stage 2 - Resume Build Pipeline

Build resume based on extracted job description and user's master data

• Craft LLM Prompt for Resume building task.
  • Persona of Job Resume Builder
  • Resume Detail Generator
  • Cover Letter Generator
• Install required dependencies: Jinja2, pdflatex, pyPDF2, etc
• Resume building module
• Cover letter generator module
• Implement relevant functionalities to support above modules
  • Text to pdf: text_to_pdf
  • logger: save_logs
  • measure_execution_time
  • JSON to latex: latex_to_pdf
  • Latex to pdf: & save_latex_as_pdf
  • Clean text from LLM: clean_string
  • Clean resume text for Latex usage: escape_for_latex
• Develop scripts to execute pipeline or workflow of resume generation.
• Format resume.tex.jinja. remove hardcoding using python-jinja syntax for JSON data.