This is a recommendation system for recruiters that estimates how new candidates suit available vacancies. It was developed for a real IT company based on their data after consulting experts. The system explores the possibility for a candidate to be: hired, rejected by the company or a candidate’s refusal.
Background
In the fall of 2016, I decided to become a data scientist.
Though I read a lot of theoretical materials, in my opinion, the best way to study something is solving real problems. I planned to implement the whole cycle from data collecting to the prediction model. As a result, I expected to get a recommendation system which will help people in decisions making.
As, at that moment of life, training period and diploma work were ahead, I decided it will help me pursue data science career.
I asked Viktor Sarapin, the CEO of V.I.Tech, if I could be a trainee at his company. He agreed and offered me a problem to solve.
The task was to create a system that could improve process of hiring employees based on information about candidates, collected over the last 4 years of company operation.
Solution
The company used software to conduct the entire hiring process. The information about a candidate was provided from the moment when a recruiter found someone till the last step of hiring. The data about a candidate includes: contacts, links to various networks, summaries, comments and characteristics added by recruiters and technical specialists.
I developed software that consolidates the data from the company system. The collected data from the system was enriched with information from external sources as GitHub and StackOverflow.
Before the data is sent to the pre-trained model, which in turn gives the probability of each of the possible statuses, it went through filtering and preparation.
- Hired
- Withdrawn
- Rejected
The results of the prediction are recorded in the database.
Now when this pipeline works, the recruiters can see a dashboard with all active candidates on it, the probability that they'll become employees and probability that candidates will pass the hiring step and go to the next one.
Conclusion and thanks
I got the highest mark with honors for the diploma project. The experience that I received was an excellent foundation for the beginning of a data scientist career.
I would like to express my gratitude to people from V.I.Tech company: Victor Sarapin - CEO, for this opportunity, to Irina Salutska — a company chief recruiter, for their assistance in understanding of business domain specifics.
Also, I want to thank Tatyana Kodliuk (data scientist) who helped me to create a pipeline and shared her great experience. Tatyana helped me to feel more confident with solving problems in the world of Data Science.
Environment
python, pandas, scikit-learn, seaborn, casperJS, phantomJS, PostgreSQL, PHP, JS, AngularJS.
Dashboard with results of AI work
Diagram of data collection process
Database for hiring recommendation system
Message for candidate
Confusion matrix
