My end of studies project i made with 3 classmates. The objective was to build a tool to generate a cover letter
that matches with job offer based on the applicant resume.
Topics covered :
- research on best LLM models
- prompt engineering for GPT
- creating a streamlit application
- data scrapping (on web for job offer,on pdf for the resume)
- generate .docx document with code
LetterCraft
details
When applying for jobs, the crucial tool is the curriculum vitae (CV).
However, diverse CV formats pose a challenge for our project, as multi-column layouts can confuse OCR software.
To address this, we opted to begin with CVs from LinkedIn, a platform providing an online version easily exportable as a PDF.
This format ensures better recognition by OCR, streamlining the extraction of context and details from the CV.
We chose to use the solution offered by OpenAI with its GPT-3 model. This approach gives us access to a powerful, high-performance model while minimising the hardware requirements on our side, and the model also has the advantage of being multilingual. It can therefore have a CV in English and an offer in French as input.
We chose to use the solution offered by OpenAI with its GPT-3 model. This approach gives us access to a powerful, high-performance model while minimising the hardware requirements on our side, and the model also has the advantage of being multilingual. It can therefore have a CV in English and an offer in French as input.
(very) quick resume of the project idea
testing meta model (ggml-alpaca-7b-q4)
project application (streamlit) with output
project run