"Employment is often considered an important indicator for immigrant integration. Entering the labour market has great significance for people moving to Finland, but also for the Finnish society and, at local level, for the municipality. Acquiring and retaining work enables immigrants to earn an income and stabilise their economic situation. It also helps them take their place in the community and the new home country: to learn the language and cultural practices, to understand how the municipality and Finnish society work, as well as how to build social relationships and networks." (http://www.kvartti.fi/en/articles/immigrants-and-employment-helsinki).
"less unemployed people there are, more wellbeing there is in the world." (track challenge)
So we decided to build an app that helps recent immigrants land to a suitable job in their new home.
What it does
A mobile application guides you through a number of flashcards with a predefined number of skills, automatically extracted one-word keywords written in your native language, or emojis. Every swipe (“this sounds OK to me!” or "Nope" encoded with emoji) updates your internal profile and helps us find a suitable job in the database. You can also unlock new jobs by taking educational courses. This is suggested right away when applicable. For instance, if you say you don't speak Finnish, language courses flashcard will follow.
Once a list of suitable job listings is created, it is shown on the screen with the details needed (location, part-time/full-time, salary, etc.). You can then apply for a job or carry on the search.
The search is not limited, so, let's say a Finnish speaking software engineer can also find a job through the app.
The app is backed with a web service which was also developed during the hackathon.
How we built it
We parsed https://paikat.te-palvelut.fi/tpt/ jobs and built an online recommender system on top of this data: we extracted keywords from raw job descriptions and matched them to job categories and jobs. This was done with a freshly baked neural net (hashtag deeplearning :). The keywords are used in the flashcards in the app.
We also extensively studied The Organisation for Economic Co-operation and Development datasets (http://www.oecd.org/, quite an interesting read!) to create an ability-to-occupation mapping model.
Courses data are taken from: https://www.hel.fi/sto/fi/opiskelu/maahanmuuttajat-immigrants/immigrants.
Service backend is rock solid AWS/php/MySQL.
Challenges we ran into
Nobody in the team speaks Finnish, so we could actually face the problem ourselves. At some point, an intermediate version of the app was often asking if a user likes to party, though no relevant jobs were eventually found.
The dataset is quite noisy so extracting keywords was a challenging task.
Accomplishments that we're proud of
The app actually works and the idea was eventually implemented though nobody has been employed yet. A unit test was written (for the first time in history of hackathons).
We are especially proud of the UI/UX - spent quite a while as we think this is important.
What we learned
During the hackathon we learned much much more about the problem from different angles and the initial project idea proved to be quite relevant (thanks our mentor for the advises). We're happy to have this solved somehow.
We also learned quite a few finnish words. Communication = Collaboration! Did you also know there is an animal sled driver job posting? How cool is that!
What's next for Path
The challenge is global - so the app can be useful in Russia and all around the globe as well. More jobs datasets can be added.