Both of us wanted to work with machine learning, and the Merit|Edge challenge seemed like a very interesting challenge.

What it does

The user inputs a folder containing any number of resumes. The user can also choose to put in a list of keywords that they want to emphasize. The program will then go through the resumes and extract features and use a neural network to determine the final score of each resume. However, the unique point of our resume parser is that it learns over time based on the user's preferences. After the program sorts the resumes based on their score, the user can click on them to view them. After viewing, they are asked input a score, and as time goes on, the program will learn the habits of the user and will conform to his specifications.

How I built it


Challenges I ran into

Feature extraction is an extremely difficult task. It is also very hard to make a good-looking GUI in Java

Accomplishments that I'm proud of

We are very proud of making the neural network and machine learning algorithms in house instead of using an already implemented one such as Microsoft Azure. This allows us to make it dynamic, adjusting to the user with every single iteration. That is what we are most proud of.

What I learned

Learned how to manipulate PDFs in Java. Learned how to work with new teammates who we had never met before. Learned that python is incredibly hard.

What's next for Perusamé

Implementing asynchronous display. It is not necessary to parse all thousands of resumes at once. As resumes are parsed, they can be displayed. This will significantly reduce runtime.

Built With

Share this project: