Degree

Bachelor of Science (Computer Science)

Department

Department of Computer Science

School

School of Mathematics and Computer Science (SMCS)

Advisor

Dr. Imran Rauf, ssistant Professor & Program Coordinator BS(CS) and PhD (CS) Programs

Co-Advisor

Nabeel Farooqui, Senior Software Engineer Folio3

Keywords

Artificial Intelligence, Recruitment Technology, Large Language Models, Retrieval-Augmented Generation, Computer Vision, Natural Language Processing, Applicant Tracking Systems

Abstract

Traditional Applicant Tracking Systems (ATS) often rely on static resume data, potentially overlooking nuanced candidate capabilities. RecruitWise addresses this limitation by introducing a dynamic, AI- powered recruitment engine designed to augment ATS functionalities with interactive, conversational interviews and robust proctoring. The primary objective is to provide a more insightful, efficient, and fair candidate screening experience. Employing an Agile methodology, RecruitWise integrates several key components: an advanced resume pre-screener that combines semantic similarity with a custom-trained resume format detection model for initial filtering; a conversational AI interviewer (leveraging LLAMA3.3 via Groq Cloud with Retrieval Augmented Generation using candidate resumes for context, and Text-to-Speech/Speech-to- Text capabilities powered by ElevenLabs); a multi-modal cheating detection system analyzing webcam footage (YOLOv8 for object detection, MediaPipe for gaze tracking) and browser activity for interview integrity; and an AI-driven scoring mechanism to evaluate interview transcripts against job-specific skills. The system is designed to conduct personalized interviews, process candidate responses in real-time, identify potential academic integrity violations, and provide structured candidate evaluations, thereby offering a significant enhancement to conventional recruitment workflows.

Tools and Technologies Used

• Frontend: React.js, Tailwind CSS • Backend: Nest.js (primary application server), FastAPI (for AI interview session management) • Databases: MongoDB (primary data store), Redis (for refresh token storage, in-memory caching) • AI - Resume Screening & ATS Integration: – Document Type Classification: Custom Python script with PyPDF2 and rule-based logic 15 – Semantic Matching: Sentence-Transformers (all-MiniLM-L6-v2), pdfplumber, NLTK • AI - Conversational Interviewer & Scoring: – LLM: LLAMA3.3 – LLM Deployment: Groq Cloud (via API) – RAG Framework: LangChain, Chroma (vector store), HuggingFaceEmbeddings • AI - Cheating Detection: – Object Detection: YOLOv8 (ultralytics library) – Gaze/Mesh Detection: MediaPipe – Processing: CUDA (for GPU acceleration on a local machine) • Speech Technologies: – Text-to-Speech (TTS): ElevenLabs API – Speech-to-Text (STT): ElevenLabs API • Cloud Storage: AWS S3 (for resume storage, webcam/screenshare footage) • Deployment & Infrastructure: – Containerization: Docker – Frontend Hosting: Vercel – Backend & Microservices Hosting: Azure Containers • Authentication & Security: JSON Web Tokens (JWT - Access Tokens, Refresh Tokens) • Browser Monitoring: JavaScript-based focus detection, active tab detection, screen detection

Methodology

TheRecruitWise project was developed using an Agile methodology. The development process was orga nized into two-week sprints, allowing for iterative progress, regular feedback incorporation, and adaptive planning. This approach was particularly beneficial for complex components such as the AI interviewer model and the resume screening module, which underwent significant development and refinement over multiple sprints. Each sprint involved planning, development, testing, and review

Document Type

Restricted Access

Submission Type

BSCS Final Year Project

Download

Share

COinS