Sieve.ai

Let Us Sort the Resumes

🚀 Runner Up at Techathon, organized by GEP during summer internship 2021

About the Project

Every day, thousands of Job Listings are created to help the companies meet the ever growing demand of market by recruiting fresh talent. The job of a Recruiter is a very hard and important. It’s like identifying a needle in a haystack. But in the age of Artificial Intelligence it can be eased out for them through our product, Sieve.ai.

Sieve.ai reduces the burden on a Recruiter by leaving the tedious and repetitive task of going through thousands of resumes, of which only are handful are relevant. Just upload all the CVs into the system and receive a list of names of candidates and their respective CVs in descending order of their relevance to the Job Description. Reduce your Recruitment Turn Around Time with just a click.

Getting Started

Start developing locally.

Step 1: Clone the repo

Fork the repository. then clone the repo locally by doing -

git clone https://github.com/rohitbakoliya/sieve.ai.git

Step 2: Install Dependencies

cd into the directory

cd sieve.ai

Install dependenices for Web-App

cd web-app/api
yarn install

cd ../client
yarn install

Install dependenices for Flask backend

cd middleware
pip install -r requirements.txt

Step 3: Setup .env

To run the server you will also need to provide the .env variables

• create a new file .env in the root
• open .env.EXAMPLE
• copy the contents and paste it to the .env with valid keys

And you are good to go

cd web-app/api
yarn dev

cd middleware
python app.py

Technologies used

• Node
• Express
• MongoDB
• Typescript
• React
• Redux
• Ant Design
• Flask
• Numpy
• Docker

Features

• 2 Layer Filter Mechanism:

  • Role-Based – It features an AI classifier trained with 900+ resumes that can classify the resumes into 20 broad categories of Job roles with an accuracy of 95%.
  • Similarity with Job Description – It extracts the keywords from the inputted job description and maps them with resume based on cosine similarity(document similarity) to get a score percentage, which is used to sort and display the best resumes on top

• It can process up to 3000 resumes per hour

• Added Google OAuth and Password-based authentication to maintain the privacy

• Recruiters can export shortlisted candidates details, scores as CSV

Architecture

• Dataset Used
• Architecture Design

Author

• Rishabh Malik
• Rohit Bakoliya
• Rohith Saji
• Sheetal Singh