Ruzi: Smarter Resume Evaluation

Our Home Page
Require User Auth to Review Resume
Sign In Page
Login Page
Simple but Efficient Design
Clear Results Displayed

All of this information is on our figma site linked below please check that out.

Inspiration

We took inspiration for this project from our own struggles in applying to jobs and internships. This was already an idea Brandon had before that we decided to bring to life.

What it does

The user uploads their resume as well as the job description of the job they are applying to. We then compare the uploaded resume to a database of applicants for similar jobs to see what those who were accepted showed on their resume that are not on the users. We run this through an NLP algorithm and a custom ml algorithm to receive an initial expected rejection or selected result which we then run through google gemini which is also added to the weighted confidence score.

How we built it

We started by designing the front-end in Figma which we then recreated in our own code. We used flask and jinja templates for the web structure and connecting to the backend. For our ML model we used trained models for each job category in our dataset averaging about an 75% accuracy when predicting acceptance rates. After adding the gemini layer the accuracy was significantly improved.

Challenges we ran into

We had to change our algorithm for predicting results multiple times. Our initial attempt was just to find skill overlaps between the job description and the resume, but this gave us about a 52% accuracy which is kind of abysmal. After trying this multiple times with different skill labels to no avail, we decided to try comparing applicants against other applicants instead of against the job application. Doing this instantly boosted our accuracy. There were also so many git merge conflicts.

Accomplishments that we're proud of

We are proud of starting and producing a solid MVP. Getting the AI to work was also a big hurdle that we were able to overcome. (Also manifesting a placement)