An IT company with large software development operations has multiple simultaneous software projects. Some projects are owned and fully run by the company itself while in most other cases it develops software for its customers. Business is constantly growing as projects scale and the number is increasing. This growth along with regular temporary labor needs for specific projects is supported by the internal HR team.

Due to the fast growth of the IT industry market dictates large and small companies must compete for talents and constantly sharpen their hiring process. Besides competition, hiring in IT has higher requirements for recruiters who also must demonstrate enough IT knowledge to be able to evaluate applicants' CVs and support screening interviews to detect and disqualify unskilled applicants.

In most cases, proper candidate evaluation requires professional analysis of sample computer code provided by the candidate. This step requires an expensive time of the senior developer and there may be hundreds of code samples. It means that resolving the shortage of software developers requires more hours of software developers for candidate evaluation. It's a vicious circle.

Vision Systems built a complex software solution that automatically evaluates job applications and computer code samples. Job applications and computer code samples are uploaded to customers' job portals (optionally link to GitHub). Recruiter accessing applications sees application and computer code rating next to each application with the ability to filter out candidates not showing the level of skill required for the position.

Recruiters' performance was increased by 2+ times as a big part of applications was filtered out after automatic evaluation. The requirements for recruiters' skills and experience were also decreased since the hiring process became less dependent on the recruiter's skills. The number of computer code samples evaluated by senior developers has decreased by 7-10 times as only samples of automatically pre-qualified candidates required verification by humans.

The core of an AI system is software that uses machine-learning algorithms. We used a large amount of historical data in combination with artificially synthesized samples to train ML models to rate candidate CVs and sample computer code. In the first stage, the system supports a few popular programming languages: PHP, JavaScript, and Laravel.

Once a job application is uploaded it is analyzed by a machine learning model. Assessment result is returned and stored in the database with the application allowing the recruiter to filter and sort candidates in accordance with their score.

This solution can evolve in multiple directions. The first obvious way to increase solution efficiency is adding more programming languages. Another way is to train system to evaluate UI/UX designers' portfolios or include candidates state of mind and emotional condition basing at the live interview.