To run app locally (requires OpenAI API key):

$ export OPENAI_API_KEY="your_api_key_here"
$ pip install -r requirements.txt
$ streamlit run app.py

To run app via Streamlit Cloud:

Go to https://candidate-recommendation.streamlit.app/

Assumptions:

• (Only if running app locally) User has an OpenAI API key.
• Resumes contain the candidate's name.
• Job description and resumes are all in English.

Approach:

1. Collecting user input
   • Job description (text)
   • Resumes (uploaded file in PDF or .txt format, or text)
     • User can select number of resumes to input manually
2. Embedding generation and similarity calculation
   • Embeddings for job description and resume text generated using SentenceTransformers all-MiniLM-L6-v2 model
   • sklearn cosine similarity function used to calculate similarity between job description and each resume
3. Displaying top candidates
   • 5 candidates with highest similarity scores are displayed in descending order by similarity
   • Candidate name is returned by OpenAI model gpt-4o-mini
   • gpt-4o-mini also generates a summary on why the candidate is a good fit for the job.

Known Issues:

• The app sporadically displays the following error:

  NotImplementedError: Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta to a different device.
  

  This issue is almost always fixed by simply reloading the page.