In order to reduce the latency in the job market, we came up with the idea of building an smart assistant that would make you discover and stay up-to-date with the job market. For this purpose, we decided start small and try to imagine a local solution that could help people keeping an eye on the Finish job market. Because there are already so many (good) platforms out there, we didn't want to come up with an extra one. In fact, we wanted the opposite, we wanted the job to come to you instead of the other way around. By thinking about it, it appeared to us that there were already a client installed on every person smartphone/computer that could solve this purpose: Facebook Messenger! We believe job-market actors like duunitori would highly benefit by adding social interaction with their users.
What it does
Job hunter is a simple interface to interact with the duunitori.fi API through Facebook messenger. He is your personal Facebook assistant who will make sure that you land your dream job.
As an user, you can:
- Start a conversation with Job Hunter on facebook Messenger
- Authenticate an account without needing to leave facebook thanks to Oauth.
- Create preferences (city+job) through the conversation
- Update/Check these preferences anytime through the conversation
- Retrive a list of jobs based on your preferences using Artificial Intelligence
- For each job, apply through the website of the provider in 1 click
- When running out of jobs, create an alert to receive an alert (notification) when a new job comes out
How we built it
- Server side is Ruby on Rails
- Client side is Facebook messenger
- Natural language processing and machine learning is made with Python
- Scheduler routines based on Cron jobs (creating matches, sending notification, polling new data from api)
Challenges we ran into
- Natural language processing on Finnish textual data
- Finnish API, we don't speak Finnish!
- No previous knowledge of Facebook developers environment
- Finding the best hyper parameters for A.I. modelling
- Make a ruby middleware interact with Python A.I. code
- Long time to define the MVP, going back and forth between two ideas (searching live vs notifications) and ending putting those two ideas in the MVP!
We're proud of
- Handle Finnish results provided by the the duunitori.fi API with Natural language processing
- Having built a featured, demoable app in 2,5 days using technologies we didn't know before
- Create a product we'd like to use if we were looking for a job in Finland
- The fun we had on the whole process on building this app
- Some Finnish words like _ ohjelmoija _
- Python natural language processing based on cluster creation was a new concept I learnt (thanks to my team-mate Omer Shafiq)
- Patience is the key
Next steps for Job Hunter?
- Create a detailed README on how to run the project on a blank environment
- Add unit tests
- Deploying to production environment
- Tracking issues on the repo to attract open-source contributors
- Building new features accessible via the messenger menu (history of research, notifications properties...)
- Handle more user preferences than city and desired positon
- Implement other job-market APIs to provide more jobs
- Add deep learning to the Job Hunter so that it naturally communicates