Candidate Ranking and Machine Learning

In the summer of 2015, I had the opportunity to develop a candidate ranking algorithm using big data, predictive analytics and machine learning. By the end of the internship I developed a web app that ranked and matched candidates for a given job description and presented it to the company.

The project required 6 programming languages, large spaces of whiteboards, Amazon ML, brainstorming with anyone & everyone, and–definitely–caffiene.

How It All Worked

First, the recruiter enters the job title and job description in the fields shown to the right. After pressing submit, a parser identifies keywords that indicate what the recruiter is looking for (i.e. location, work status, skills, etc.).

The identified data picked up by the parser is then placed into a user-friendly form. In which the recruiter can attribute weightages, add additional fields or remove extraneous skills.

After pressing "Get Results" the algorithm is run; an ideal score is created based on the user weightages and run against other candidates. The lists and ranks of all the candidates that meet the base criteria are then presented. The result screens are shown below.

Scroll down if you'd like to read about the algorithm.

Developing this algorithm took almost 6 weeks of brainstorming.
While I would love to share everything with you, there are only certain details about the algorithm I can share.

The following section will discuss the algorithm.