Service:

Development of an AI model

Stack:

Python, Pytorch

Development period:

11 months

Today we will tell you about our interesting startup that our team tried to implement in 2019.

Recrut.ink is a unique recruitment service based on artificial intelligence, taking into account the psychotype of the manager.

This project is based on a scientific approach using methods for diagnosing stable psychological personality traits, such as the "Big Five" and Raymond Cattell's 16-factor questionnaire.

There is a problem in the world with regard to the recruitment of personnel for companies. These problems are confirmed by statistics provided by GlassDoor.

According to the study:

• On average, the HR department needs 83.3 hours to analyze the resume of one candidate.
• Viewing one resume takes only 5-6 seconds.
• More than 250 calls will be required to conduct 4-6 interviews.
• More than half of the funds spent on recruiting go to recruiters, despite the fact that 65% of resumes are ignored.

In addition, research results from Monster, Projecttimes and BLS say:

• 89% of employers inspect candidates' profiles on social networks before inviting them for an interview.
• The success of projects in 60-70% of cases depends on time and resource management (PCPM).
• Every fifth employee leaves the company due to various psychological aspects of interaction with the boss

Hence, our team came up with a solution to this problem:

Eliminate the human factor using the AI model in order to automatically invite candidates for an interview and select specialists in the psychotype of the head.

At the same time, so that the relevance of the selection of an employee is more than 90%

• Reduction of financial costs for recruiters and labor costs for the selection of a quality employee by at least 90%
• Improving team efficiency by 30-40%

• The service automatically parses the job description and determines the psychotype of the manager

• Selects the best resumes from millions

• Automatically determines the psychotype of a person by profile in social networks

• Looks through the manager's Google calendar and sends an invitation

• In case of confirmation, puts the meeting in the manager's calendar

To avoid long psychological testing, the AI model should analyze the psychological profile of the candidate, according to the social network profile provided by him, from the following possible options: LinkedIn, Facebook, Twitter.

We have outlined the tasks that were required to implement functionality of a similar scale:

• A model based on the Kettel methodology, with an emphasis on the Lexical Hypothesis.
• Collecting data from users: information from social networks, resume, TIPI test.
• Using a semantic differential to understand job expectations.
• Creating a psychological profile of a candidate based on data from TIPI, semantic differential and Cambridge Analytics models.
• Selection of resumes and employees in accordance with the requirements and personality traits.

Collecting and analyzing data about candidates:

First of all, we get information from social networks and candidates' resumes, then we analyze their personality traits using the TIPI test and the Charles Osgood semantic differential method.

Accounting and creation of a psychological profile:

Based on the results of the analysis, our AI model creates a psychological profile for each candidate, taking into account both their personality traits and job expectations.

Selection and combination:

Further, comparing the psychological profile with the requirements of the vacancy, we select recommendations based on comparison, and, taking into account both professional skills and personal aspects, we find the best combination of the candidate and the project.

Analysis of the candidate by the Kettel method

When developing the artificial intelligence model, we relied on the methodology of Raymond Kettel, namely the Lexical Hypothesis, according to which the most important individual differences are manifested in the concepts of language.

Training and operation of the AI model:

After that, we trained the AI model so that it analyzes text data from social networks according to this methodology.

Which included highlighting keywords and concepts related to individual traits.

Result

As a result, an online platform with our integrated AI model was to be implemented, which outputs the candidate's psychological profile based on the following data: TIPI, semantic differential, Cambridge Analytics models.

Thanks to this, the AI could not only be able to pick up a resume for a vacancy, but also pick up an employee for a manager, reducing the percentage of specialists leaving and increasing the friendliness of the atmosphere in the team.

But due to the end of funding, this project is temporarily frozen!

Friends, if you want to develop or integrate your own AI model, you can contact us on the website.