About the Project

Nexrecruit AI is a next generation Recruitment Platform on a mission to ensure that all recruiters are able to build an A team with minimal efforts. CV Evaluation is normally a lengthy process that requires ~5 minutes per CV, spent in reading, highlighting, and grading. Preparing the interview shortlist basis these notes includes multiple rounds of iterations from other recruiters as well.

The shortlisting process is rarely standardized across recruiters, and may lead to missing out on capable candidates due to oversights, knowledge gaps, or biases.

Nexrecruit AI wanted to solve this challenge.

Key Challenges

[1] CVs are generally submitted by the candidates in multiple file formats, including images, PDFs, Word Docs, etc. and are spread across multiple channels like LinkedIn, Naukri, Email, WhatsApp, etc. Consolidating this information and ensuring that the same solution is able to process all file formats was a challenge.

[2] CVs are also written in multiple languages, lingos, and use different names for the same terminology. This information needs to be standardized in order to be evaluated fairly.

[3] Each recruiter may have a different philosophy of gradation, and each role may demand different skillsets to have different weightages. Hence, a one size fits all product does not work.

Approach

[1] We used Textract and DocX conversion libraries to ensure that text could be extracted from all file formats, no matter how different.

[2] Next, we used GPT 4o heavily to convert output of textract into a JSON, standardize terminologies, and grade skills

[3] We allowed recruiters to configure the peaks and troughs in the resume by allowing them to add weightages for keywords. This now meant that Java as a skill won't be considered vital for an SDE 3 role, but will be vital for an SDE 1.

Here is how the product looks like: