Textual classification of internship titles

Development of an API to classify internships (internship accepted or not by the school) given their titles. 3 approaches:

• Machine Learning : static embedding + classification algorithm
• Classification with CamemBERT
• Few-Shot Learning using a generative model (Mistral)

To install the necessary libraries:

pip install uv
uv pip install -r requirements.txt

For the Zero-Shot and Few-Shot learning approches you need to provide a HuggingFace token for the model used after doing:

huggingface-cli login

You can verify that the GPU is used with:

nvidia-smi

To run the API:

uvicorn main:app --reload

Further analysis on rejected internship subjects is also given in the notebooks (topic modeling with LDA and NMF, and clustering with BERTopic).