Profile Builder - Land your dream job(s)

Its your last year at school and you're applying for jobs. You find a role that interests you but the JD mentions technologies and problem spaces you haven't worked with before. Enter profile builder.

Comment

Inspiration

We came here with the intention of learning something that should help us in our career. Then we ended up making something that will help us all in our careers (I guess that qualifies this project for social impact, category idk). Well, the problem is its too late to learn a stack (Kafka, Flink?) or about a problem space (search, recommendations) when the "applying to jobs" season is on. We wanted to build an application that looks at the skills you have, a bunch (50, 500?) job descriptions of the roles that you want to apply to in the future and suggests you skills to learn.

What it does

Given a resume, and the job descriptions of the roles the user wishes to apply, we recommend/rank the most relevant positions for the user, the most relevant skills, the ones that are missing skills and the projects one can do to make up for them.

How we built it

Sugar, spice, and everything nice! The core of our app is information retrieval, and semantic similarity. We used TF-IDF for computing the relevance between resumes and job descriptions. KeyBERT and RAKE was used to understand important keyphrases and keywords from the resume and job descriptions. We then computed semantic similarity between the two to evaluate what skills the user has and what are missing. And finally, we used the missing skills as a search query to recommend/rank projects the user should do.

Challenges we ran into

  1. Finding the datasets for training and evaluation.
  2. Machine learning only works for others.
  3. Developing a user interface.
  4. Time

Accomplishments that we're proud of

  1. The core algorithm does work better than we hoped it to be. Remembering the popular saying inaccurately, all models are bad but some are useful.
  2. We were able to complete most parts of it by pruning the other complex parts, like UI and Django backend, utilities, etc.
  3. This tool could really benefit students and professionals likewise, in searching for their dream jobs (if built properly :P).

What we learned

A lot of things. Django, HTML/CSS, Python, IR, Semantic Similarity, Ranking but most importantly how to beat linkedin at recommendations.

What's next for Profile Builder - Land your dream job(s)

Landing our dream jobs!

Built With

Share this project:

Updates