How Vector Search Can Help You Find the Perfect Candidate

Saurabh Rai
5 min readJul 19, 2023

--

Photo by Christina @ wocintechchat.com on Unsplash

When searching for a perfect candidate for a role. There are many factors one has to consider. And even before the resume reaches the recruiter's desk. It has to go through a lot of checks. Usually, an ATS (Applicant Tracking System) parses the resume and then filters based on keywords. And even then, recruiters don’t get much time to read through the resumes, popular internet saying goes. Recruiters spend an average of 20 seconds checking whether the candidate fits the role. This approach is flawed because resumes represent less than a candidate’s experience.

This article explores the problems in recruiting and how vector databases can solve those problems. And if you’re a search enthusiast working on Vector databases, semantic search, or a technical recruiter who wants to know cutting-edge technology. Read along, and we’ll solve a crucial problem together.

Identifying the Problems in the Current Recruiting Condition.

Resumes are multifaceted, representing a candidate’s work experience, personal projects, technical skills, and other experiences such as volunteering, open-source projects, blogs, and the hard work they’ve put into their career. It’s also a candidate's desire to express his/her/their interest in joining a company’s team and valuing them.

A resume represents more than a candidate's work experience.

This problem of spending time on a candidate’s resume can be solved using vector search and vector databases. It’s the perfect tool for finding the best talent. First, let’s understand how vector search works and its importance.

How Vector Search Works?

Vector search transforms data (like text, images, or sounds) into a mathematical representation called a vector, which is a list of numbers. These vectors are stored in a database. When a search query is entered, it’s also transformed into a vector using the same method. The Vector Database Engine then searches the database to find vectors that are most similar to the search query’s vector. This similarity is determined by calculating the distance between the vectors. The Distances used can vary depending on the Vector Database or the algorithm. Common ones are Cosine, Dot Product, Euclidean, Manhattan, Hamming, etc. The vectors with the shortest distances are considered the best matches to the query. This makes vector search efficient and accurate at retrieving relevant results, even for complex or vague questions.

Why is Vector Search Important in Recruiting?

The human mind is biased. Even when deliberately trying to find out ways to reduce the biases. As humans, we face blind spots in our decision-making. This transfers to how we perceive a candidate: gender biases, race, religion, culture, and country. We can be more selective towards people who reflect our surroundings or mindset. And to remove this bias. Vector Search can be the perfect tool.

Since all the resumes are stored as numbers or vectors, the algorithm treats the search query, which is the job description. And matches it with the most relevant resume vectors. This produces fast results which are unbiased and represent candidates who align well with the job description overall rather than focusing on certain keywords.

Transforming Recruiting with Vector Search

And Vector searches don’t discriminate since everything to it is just numbers. So, you can get a talent pool of amazing people from diverse backgrounds who can come together to solve various problems and provide creative solutions.

How Can You Create a Recruiting Solution Using Vector Databases or Vector Search Engines?

There are many Vector Databases available in the market. And many of them are open-sourced and free to use. There’s also a tool called Resume Matcher, which is also open-sourced and free to use.

Let’s see how many vector databases there in the market are:

You can read more about them in this article by Dmitry Kan. There are more databases than the ones mentioned above, but they’re enough for our use case to get started.

How to make your resumes work with Vector Databases?

This is as simple as it sounds. With the easy-to-use APIs, documentation, and examples available to us. All I can say (in the scope of this article) is:

  1. Use a script to parse the set of resumes.
  2. Send the resumes to a vector database.
  3. Send in your search query.
  4. Let the magic ✨happen.
  5. You will get the top answers based on what job description you sent in.

While this needs an article of its own (coming soon), you can also use Resume Matcher, which uses Qdrant to solve the same problem. And it works both ways for candidates to improve their resumes. And recruiters who are looking for better candidates.

Vector Search is the best tool for a recruiter.

Try this approach, and you will surely see amazing results. Vector Databases are a powerful tool. And they can solve a lot of problems we face. This is just a small example.

Summarizing what we just read

Using Vector Databases & Vector Search in recruitment revolutionizes the conventional and sometimes inaccurate process, making it more efficient, unbiased, and simplified. The system transforms resumes and job descriptions into mathematical vectors, enabling it to match candidates to job roles based on multiple factors, not just keywords.

This enables better candidate selection, and the whole process can be automated so that whenever you get the best candidate, you can interview them and hire them.

If you liked this article, share it with your peers, team, or on the internet. You can follow me for more such articles on Medium. Here’s my LinkedIn and GitHub. Say Hi, and let’s connect!

--

--

Saurabh Rai

Software Developer | Open Source | Building resumematcher.fyi | Working on AI Search