DoCer

We didn't know how to win ICHack, so we decided to predict it.

Comment

Inspiration

We didn't know how to choose a winning idea, so we decided to predict it. Sponsors no fear! For this ICHack are lifting the burden of judging submissions by leveraging Data Science in a fun and creative way. We are going to use tinder to predict the 2022 ICHack winner (besides us)!

What it does

To understand what makes a winning team, we talked with sponsors themselves and generated three main factors:

  1. Technical skills
  2. Group working
  3. Accuracy with the task Our platforms generates metrics about each of the above points for the teams that agreed to provide their data. It then combines that data with a weighted sum, that respect how much sponsors weight each bullet point.

How we built it

1. Technical Skills

We assess the technical skills of each group member by parsing their CV with NLP algorithms and getting insights about Work Experiences, Individual Projects, Tech Stack and General Skills. From the CV scraping, we also retrieve the GitHub username, with which we look for lines of code, repositories contributed to, contributions and stars to own repos and we validate the Tech Stack. Each of this data, is then standardised dividing it by the mean of the entire dataset for the specific entry.

2. Group Working

We look for how diversified the skillset is, as well as the technical combined potential. For the former, we compare each team member's Tech Stack data generated in part (1). For the latter, we just multiply the tech potential indicator generated in (1) with the number of members

3. Accuracy with the task

In order to win, the team has to have the personalty and creativity the sponsors are looking for. We therefore created a fun tinder-like personality test, that models the 16-personality trait concept, but with funny memes. We asked both hackers and sponsors to do the test, from which we compare the answers, to see how much their coder personalities reflect. We also added a pinch of fun :)

Challenges we ran into

  • The main challenge was getting people to submit their CV and spend precious Hacking time to make our personality test. For this reason, we received few CV submissions, and were not able to implement our prediction on the whole ICHack scale. Also we couldn't force people to submit their CVs, so for some entries we found ourselves without one from which to assess the team's coding potential.
  • We also were only a two people team, building a microservice platform leveraging APIs and NLP, which resulted in long hacky hours.
  • Finally, a fun challenge was deciding the proper heuristics to use, we overcame that by just asking to the people that are really judging what they value.

Accomplishments that we're proud of

We are proud of the sleek UI/UX of the Web App. We think the app is the perfect mixture of gathering valuable data in a fun way. We are also proud of having build a Microservice in Python for parsing CVs and Scraping GitHub profiles, and the tests we carried on and the results we received show that the integration works as expected. But most of all we are proud of our personality test platform logo! DoCer is the perfect tind*er-like **docker* application developed for DoCSoc ICHack!

What we learned

By working in a Tech-diversified team, we learned from each other. For Robbie it was the first time making a full stack Web App, as well as setting up Microservices with Flask in Python. For Filo, it was the first time implementing a moderate NLP application. In any case, developing such interesting tech in a funny application has only made it more entertaining, which is what we believe Hackatons should be!

Share this project:

Updates