Flas app to predict the level of seniority using unsupervised machine learning techniques.
- Scrapped publicly available employment history data
- Processed the data using several word embeddings, e.g. Spacy, FastText, GloVe and BERT
- Implemented different ranking models using PyTorch: RankNet, ListNet and my own approaches
- Used bubble-sort distance to evaluate different models' performance
- The best model (RankNet + FastText) resulted in 87% sorting accuracy on validation sample
- Created an app on Flask and deployed it to Heroku