Jobalytics: Using Machine Learning to Perfect Your Resume
Your resume has six seconds to grab the recruiter's attention. Make it count.
UWaterloo ENGHACK18 project developed initially in 24 hours, using pre-trained NLP algorithms/APIs to perform transfer learning in Jupyter Notebook to build a tailored resume screening web application.
- Job Match
- Adjusted Word2Vec model to classify based off keywords w/ a sieve to catch blacklisted words
- What Works & What Doesn't
- Adjusted unsupervised neural network to cluster based on word morphology
- Frequent Words
- Built a tf-idf model to perform k means clustering
- Personality Radials Classification
- Regularized algorithm to correct for high variance/overfitting
Server runs on the Express framework in Node.JS with ML algorithms called upon through python-shell. The 6 pronged analysis is visualized using Wolfram One's cloud computation platform for a friendly user experience. The front end design was inspired by the interactivity of Google's Cardview template.
Potential Future Additions
- Sentiment analysis to better align with company culture
- Facial recognition to identify emotion in interviews
- MongoDB to allow for multiple resumes/account
- JSON Web Tokens to form secure resume editing groups