HireHub — AI-Powered Enterprise Interview Simulation Platform

Built for Hack & Forge 2026 — Problem Statement 2: Web-Based Intelligent Selector–Applicant Simulation Platform for Enterprise Recruitment

---

What is HireHub?

HireHub is an intelligent, fully automated interview simulation platform that conducts structured enterprise interviews, adapts questions in real time based on candidate performance, and generates objective suitability reports — without any human interviewer involvement.

The system uses a multi-agent AI pipeline grounded in company documents via RAG (Retrieval-Augmented Generation), ensuring every question asked and every answer scored is traceable back to real organisational knowledge rather than generic AI output.

---

Problem It Solves

Problem	How InterviewIQ addresses it
Interviewer bias	Fully automated scoring — no human subjectivity
Inconsistent questions	AI generates role-specific questions from the same JD every time
No objectivity	Every answer scored on 4 dimensions against company documents
Slow hiring pipeline	Instant suitability report generated at session end

---

Live Demo

Repo: https://github.com/SwarnabhG07/Project-Rejection

---

Features

Done

• Candidate landing page and dashboard
• Resume upload with PDF text extraction
• Proctored examination session — webcam active, fullscreen enforced, tab-switch detection
• Real-time speech-to-text answer capture using Web Speech API
• AI-based adaptive questioning — questions generated from company documents
• Answer scoring system — Accuracy, Depth, Clarity, Relevance (0–10 each)
• Suitability report generation at session end

Planned

• Multi-language interview support
• Video proctoring with facial recognition and gaze tracking
• ATS (Applicant Tracking System) integration
• Interviewer dashboard with session replay
• Industry-specific question libraries (engineering, finance, operations)
• Real-time interviewer coaching panel

---

Architecture

Multi-Agent Pipeline

Inputs: Required Skills + Company Documents + Candidate Resume
            |
        Agent 1 — Topic Extractor
        Compares resume against Required skills.
        Extracts all relevant topics and checks,
        whether user is fit for the job
            |
        Agent 2 — Session Orchestrator
        Chooses what topic to ask question from,
        and frames a query which is used to get
        context from a RAG architecture for framing question.
            |
        Agent 3 — Question Framer
        Based on the context, generates questions
        which the user has to answer.
            |
        Agent 4 - Evaluator
        Scores answer against retrieved context on more dimensions.
        Writes to session scores and reason for the score into a file.

RAG Pipeline

Company Documents
      |
  Chunking (256–512 tokens, 50-token overlap)
      |
  Embedding Model (sentence-transformers)
      |
  Vector Store (FAISS)
      |
  Query at inference time → Top-k chunks retrieved
      |
  Chunks passed to Agent 3 (question framing) and Agent 4 (evaluation)

---

Tech Stack

Layer	Technology
Frontend	HTML, CSS, JavaScript, Web Speech API
Backend API	FastAPI, Python
Authentication	FastAPI sessions
PDF Processing	pdfplumber
AI / LLM	Gemini API via Chat Completions
NLP Evaluation	sentence-transformers
Vector Store	FAISS
Session State	Python dictionary (single user demo)

---

Project Structure

Project-Rejection/
├── app.py                  # FastAPI application — all routes
├── static/
│   ├── style.css           # Main dashboard styles
│   ├── styleexam.css       # Exam page styles
│   └── js/
│       ├── script.js       # Dashboard JS — upload, skills, webcam
│       └── scriptexam.js   # Exam JS — speech-to-text, proctoring
├── templates/
│   ├── index.html          # Candidate dashboard
│   └── exam.html           # Proctored exam interface
├── uploads/                # Uploaded resumes and company docs
├── main.py
├── rag_engine.py
└── requirements.txt

---

API Endpoints

Method	Endpoint	Description	Auth
GET	/	Candidate dashboard	—
POST	/	Upload and extract resume PDF	—
GET	/exam	Proctored exam page	—
POST	/submit-answer	Submit candidate answer for scoring	—
POST	/save-skills	Save required skills for a role	—
POST	/upload-company-doc	Upload company knowledge document	—

---

Setup and Installation

Prerequisites

• Python 3.10+
• pip

Install dependencies

pip install -r requirements.txt

Run the server

uvicorn main:app --reload

wait for few seconds...

Open in browser

http://127.0.0.1:8000

---

Requirements

fastapi
uvicorn
pdfplumber
python-multipart
sentence-transformers
langchain
openai
faiss-cpu
redis
reportlab

---

How It Works — Candidate Flow

1. Candidate visits the dashboard and uploads their resume
2. Clicks Start Exam — webcam activates, fullscreen enforced
3. AI generates the first question based on resume + company document
4. Candidate speaks their answer — speech-to-text transcribes in real time
5. Candidate clicks Submit — answer is scored by the evaluator agent
6. System decides: follow-up question or next topic based on score
7. Interview completes — suitability report generated instantly

---

Scalability

The architecture is designed for horizontal scaling from day one:

• Stateless FastAPI — any worker handles any request, load balancer ready
• Redis session state — all agents share one source of truth
• FAISS → Pinecone swap for production vector store with concurrent access
• Agent-level scaling — each agent can be deployed as an independent worker
• Celery async tasks — score writes never block the live interview flow

---

Team

Member	Role
Swarnabh Guha	Frontend development, JavaScript, UI/UX
Aadity Setu	APIs, backend development, deployment
Ashwath Srinivasan	RAG pipeline, AI integration, system architecture

---

Hackathon

Event: Hack & Forge 2026 Problem Statement: PS-2 — Web-Based Intelligent Selector–Applicant Simulation Platform for Enterprise Recruitment Track: Enterprise AI

---