Syft: GPT Powered Resume and Job Decription Parsing System

Jun 11, 2023

syft-app
syft-app
syft-app

Image by (pexels)

Client


Syft is a UK based start-up on a mission to 'derisk' hiring and support recruiters to hire better and faster so that they can get back to the enjoyable part of theirs jobs; the part where we meet people, develop relationships, and see their recommendations thrive.


Background


The recruitment process is inherently a high-stakes game of risk, similar to placing a bet on the future performance of a new hire. The quality and effectiveness of the hiring decision largely hinge on the approach taken to review candidate resumes (CVs). Methods range from a high-risk strategy of random selection to a meticulous, in-depth analysis of each CV. This detailed review includes scrutinising every word, exploring all provided links, and gathering as much information as possible to thoroughly assess or eliminate each candidate.

A key insight in recruitment is the direct correlation between the volume of CVs reviewed and the likelihood of making an exceptional hire. However, it is common for a single job posting to get over thousand of resume from candidates.


Solution


Syft worked in partnership with Seeai to build a product that can address the challenges that recruiters face.

Key technical challenges of Syft includes:

  • Parsing and Standardising Any Resumes. As resumes come in diverse range of format, structure, word choice, it is usually difficult to develop a parser that is able to handle resume without no-prior assumption on the format.

  • Matching Job Description to Resumes. Given a job description, we want to automatically find a resume that best 'matches' the requirements provided by the job description.

  • Find Similar Resume (Resume to Resume Matching). Given a resume of an ideal candidate, we want to find a resume that is the most similar to the resume from a pile of resumes automatically.


Seeai effectively utilised document processing capabilities of Large Language Models (LLMs) to solve these challenges. Seeai trained LLM as resume and job description parser and standardiser that converts the documents into an analysis-ready format that can be stored and indexed in a database.


We pipelined LLM with Atlas Search by MongoDB to further improve user experience. Using existing search engine meant we can utilise techniques such as fuzzy search, custom scoring, geospatial-aware search, and speed optimisation.


Each recruiter has their own favourite Applicant Tracking System (ATS) of choice and Syft's vision was to build a product that can seamlessly integrate to their day-to-day workflow, rather than introducing a completely new one. For this reason, Syft was wrapped as a Chrome Extension so that recruiters can use Syft seamlessly. Seeai used open-source web application framework Vue.js to create the frontend of the application.


DEMO




Tech Stack Used


Backend

  • Language: Python

  • Framework: FastAPI

  • Async task queue system: Celery


Frontend

  • Language: TypeScript

  • Framework: Vue.js

  • Chrome exteion


Authentication

  • Firebase


Database

  • FIrebase realtime store

  • MongoDB



Access Syft here

Ready to get started?