Vishleshan — Multi-Agent Recruitment Intelligence Platform

Vishleshan — Multi-Agent Recruitment Intelligence Platform

A multi-agent AI system for semantic resume parsing, candidate matching, fraud detection, and recruitment automation — built for enterprise HR teams and developer integrations.

Architecture • Agents • Features • Quick Start • Admin Panel • Developer Portal • API Reference • Security • License

Overview

Vishleshan is a production-grade, multi-agent AI platform that automates the entire recruitment pipeline — from resume ingestion and skill extraction to candidate ranking, AI-powered interviews, and fraud detection. It uses a coordinated system of specialized LLM agents, vector databases, and asynchronous workers to transform unstructured documents into actionable intelligence.

The platform serves four distinct user groups:

Recruiters — A full-featured Applicant Tracking System (ATS) with AI screening, matching, and analytics
Job Seekers — A job discovery portal with AI resume builder, safety verification, and legitimacy checks
Developers — A SaaS API portal with subscription billing, rate limiting, and usage dashboards
Administrators — A dedicated moderation dashboard with ban/unban controls, support ticket management, and audit logging

Multi-Agent System

Vishleshan's core intelligence is powered by a coordinated system of 12+ specialized AI agents, each responsible for a distinct task in the recruitment pipeline:

Agent	File	Responsibility
Resume Parsing Agent	`agents/parsing_agent.py`	Extracts structured data (skills, experience, education, projects) from PDF/DOCX/TXT files using LLM-guided extraction
Advanced ATS Parser	`agents/advanced_ats_parsing_agent.py`	Deep ATS compatibility analysis with section-by-section scoring
ATS Compatibility Agent	`agents/ats_compatibility_agent.py`	Scores resume formatting, keywords, and structure against industry ATS standards
Skill Normalization Agent	`agents/normalization_agent.py`	Maps raw extracted skills to a canonical taxonomy of 1,000+ technical and soft skills with synonym resolution
Matching Agent	`agents/matching_agent.py`	Computes semantic match scores between candidate profiles and job descriptions using vector similarity and weighted criteria
Inference Agent	`agents/inference_agent.py`	Infers missing candidate attributes (seniority level, role category, domain expertise) from available resume data
AI Chatbot Agent	`agents/chatbot_agent.py`	Powers natural language candidate queries — ask questions like "Find Python developers with 3+ years in fintech"
Fraud Detection Agent	`agents/fraud_agent.py`	Scans resumes and job postings for plagiarism, AI-generated content, ATS keyword stuffing, and phishing patterns
Interview Agent	`agents/interview_agent.py`	Conducts AI-powered mock interviews with dynamic follow-up questions
Resume Enhancer Agent	`agents/resume_enhancer_agent.py`	Suggests improvements to resume content, formatting, and keyword optimization
Salary Prediction Agent	`agents/salary_prediction_agent.py`	Predicts compensation ranges using ML models and market data
Cover Letter Agent	`agents/cover_letter_agent.py`	Generates tailored cover letters based on resume and job description
MCQ Paper Parser	`agents/mcq_paper_parser_agent.py`	Auto-generates MCQ assessments matching job requirements
Resume Quality Agent	`agents/resume_quality_agent.py`	Evaluates overall resume quality with actionable improvement scores
Job Recommendation Agent	`agents/job_recommendation_agent.py`	Recommends relevant jobs based on seeker profile and preferences
JD Generator Agent	`agents/jd_generator_agent.py`	Generates optimized job descriptions from role requirements
LLM Router	`agents/llm.py`	Manages API key rotation across multiple Gemini keys with automatic failover, rate-limit recovery, and load balancing

All agents communicate through the RotateLLMClient, which distributes requests across a pool of API keys and automatically handles quota exhaustion, retries, and model selection.

Features

Recruiter Dashboard (ATS)

Batch Resume Upload — Drag-and-drop PDF/DOCX files with async Celery processing
AI Skill Extraction — Multi-agent pipeline extracts skills, experience, contact details, and projects
Semantic Job Matching — Vector-based scoring against job descriptions with configurable skill weights
AI Chatbot — Query your candidate pool in natural language
Session Management — Multi-round hiring workflows with configurable evaluation criteria
Gmail & Google Drive Sync — Import resumes directly from email attachments and cloud folders
Company Settings — Upload company logos, configure branding, and manage team profiles
Analytics Dashboard — Hiring velocity, pipeline health, and candidate quality metrics

Fraud Detection & Protection

Resume Authenticity Scanning — Detects AI-generated content, plagiarism, and invisible keyword stuffing
Job Posting Verification — Flags phishing scams, ghost job indicators, and clone copy-paste listings
AI Fake Job Detection System — Fully integrated 6-point verification checklist:
1. Official Website Validation (corporate domain and security certificates)
2. Recruiter Email Domain Verification (matches registered corporate domain)
3. Salary Realism Evaluation (compensation vs. local market standards)
4. LinkedIn Company Presence Check (professional networking platform lookup)
5. Suspicious/Copied Description Analysis (boilerplate scam template detection)
6. Duplicate/Repeated Posting Detection (mass automation pattern scanning)
Legitimacy Score Output — Real-time Trust Score (/100), Risk Level, Verified Company status, and Approved/Suspicious classification
Scan History — Full audit trail with dynamic interactive breakdowns

Job Seeker Portal

Dedicated Seeker Accounts — Separate login/registration with OTP verification (SMS via 2Factor + Email via Brevo)
AI Resume Builder & Editor — 7 professional templates (Modern, Classic, Minimal, Executive, Creative, Compact, ATS Optimized)
High-Fidelity Template Previews — Exact rendered images matching actual template structures
Dynamic Column Selector — Toggle between 1-column and 2-column layouts
ReportLab PDF Exporter — Multi-page 2-column PDF generation with dynamic pagination
Profile Auto-Sync — Setting a resume as "Active" syncs skills, experience, and education to the seeker profile
Double-Input Job Search — Keyword + location search with Indian city-to-state autocomplete
Real-time Notifications — In-app notification center and email alerts on application status changes
Hiring Safety Checker — Company domain authenticity scanner before applying
Salary Trends & Analytics — Interactive sector salary and demand charts
Google & GitHub OAuth — One-click social sign-in across all portals
Mock Interview Practice — AI-powered voice/chat interview simulation with scoring and feedback
Company Discovery — Browse, follow, and explore company profiles and open positions
Applications Pipeline — Track application status with visual pipeline stages

Admin Dashboard

Dedicated Admin Login — Separate /admin/login page with Combined (IP + Email) rate limiting
User Moderation — Ban/unban any seeker, recruiter, or developer with atomic Redis cache invalidation
Self-Ban Prevention — Admin email matching blocks accidental self-banning
Support Ticket Management — View, resolve, and respond to user support requests
Audit Logging — Every ban/unban action is recorded in AdminBanLog with timestamp and admin email
Banned User Enforcement — Banned users are immediately blocked on all API calls (Redis-cached, DB-fallback)
Session Invalidation — Active JWT tokens are rejected mid-session when a user is banned
Short Token Expiry — Admin JWT tokens expire in 20 minutes

Developer Portal (SaaS API)

REST API — Programmatic access to resume parsing, matching, chatbot, and fraud scanning
API Key Management — Generate, rotate, and revoke production/test keys
Subscription Billing — Razorpay-integrated plans (Free, Starter, Business, Enterprise) with monthly quotas
Rate Limiting — Redis-backed per-key monthly quotas for parse, match, chat, and scan operations
Usage Analytics — Real-time traffic charts, endpoint latency, and monthly usage breakdowns
Webhooks — Configure HTTP callbacks for async parsing completion events
Embed Widget — Generate secure tokens to mount Vishleshan UI in external applications
API Documentation — Interactive playground with request/response examples

Candidate Assessment Rounds

MCQ Round — Auto-generated Technical, Conceptual, and Aptitude questions matching job requirements
Coding Round — Custom programming challenges with in-browser code editor and test-case grading
AI Interview Round — Dynamic conversational interview with sequential follow-up questions, scoring, and feedback

Architecture

The platform is built as a monorepo with two primary pillars:

`backend/` — The AI Engine

Framework: Django 5 + Django REST
Database: PostgreSQL (Neon serverless / Supabase)
Cache & Rate Limiting: Redis (ban status caching, JWT blacklisting, rate limiting)
Vector Store: ChromaDB for semantic similarity search
Task Queue: Celery with threaded pool for async resume processing
LLM Provider: Google Gemini (multi-key rotation with automatic failover)
Authentication: JWT tokens (recruiter/seeker/admin) + API keys (developer)
SMS: 2Factor Gateway for OTP delivery
Email: Brevo (Sendinblue) for transactional emails, verification codes, and notifications

`frontend/` — The React SPA

Framework: React 18 + Vite 8
Styling: Tailwind CSS 3
State Management: Zustand
Data Fetching: TanStack React Query
Charts: Recharts
Animations: Framer Motion + GSAP
UI Components: Radix UI primitives
Routing: React Router v7 with nested layouts

The frontend hosts all four portals as a single SPA with path-based routing:

/ — Recruiter landing page
/dashboard/* — Recruiter ATS dashboard
/jobs/* — Job seeker portal
/developer/* — Developer API portal
/admin/* — Admin moderation dashboard

Platform Highlights

1. Interactive Recruiter ATS (Applicant Tracking System)

Complete AI evaluation panels with semantic match scoring, candidate filtering, and multi-round hiring pipelines.

2. Recruiter Recruitment Sessions

Track and manage all hiring rounds, candidates, and job applications in one visual board.

3. Interactive AI Resume Builder & Editor

Build your resume using high-fidelity templates with a real-time side-by-side ATS compatibility score and AI suggestions.

4. Job Seeker Dashboard & ATS Scoring

Track your applications pipeline and review real-time feedback with a breakdown of keywords, skills, formatting, and experience.

5. Developer API Portal (SaaS Dashboard)

Full SaaS portal for third-party integrations with usage analytics, Razorpay subscription billing, and interactive documentation.

6. Job Seeker Landing & Smart Job Discovery

City-aware job search, live state-to-city Indian autocomplete suggestions, and domain safety verification.

7. Market Trends & Salary Insights

Interactive wage trajectories, hiring velocity index, and region-wise job opening distribution charts.

Quick Start

One-Click Development Start (Windows)

For Windows environments, you can boot the entire local workspace (Vite Frontend, Django Backend, Redis, and Celery worker) using the provided batch script:

run.bat

Manual Setup & Requirements

1. Requirements

Node.js v18+
Python v3.10+
PostgreSQL, Redis, ChromaDB (running locally or remotely)

2. Backend Setup

cd backend
cp .env.example .env

# Install Python dependencies
pip install -r requirements.txt

# Configure environment variables in .env:
# GEMINI_API_KEYS — comma-separated list of Gemini API keys
# GEMINI_MODEL — e.g. gemini-2.5-flash
# GOOGLE_OAUTH_CLIENT_ID, GOOGLE_OAUTH_CLIENT_SECRET — Google OAuth
# GITHUB_CLIENT_ID, GITHUB_CLIENT_SECRET — GitHub OAuth
# BREVO_API_KEY — Email & SMS service
# TWOFACTOR_API_KEY — SMS OTP gateway
# RAZORPAY_KEY_ID, RAZORPAY_KEY_SECRET — Payment gateway
# ADMIN_EMAIL, ADMIN_PASSWORD — Admin panel credentials
# DATABASE_URL — PostgreSQL connection string
# REDIS_URL — Redis connection string

# Run Database Migrations
python manage.py migrate

# Seed the Skill Taxonomy table (one-time)
python manage.py seed_skills

# Run the Django Dev Server
python manage.py runserver 8000

# Run the Celery Worker (multi-threaded pool on Windows)
celery -A workers.celery_worker worker --loglevel=info --pool=threads --concurrency=4

3. Frontend Setup

cd frontend
npm install
cp .env.local.example .env.local
# Configure VITE_GOOGLE_CLIENT_ID, VITE_GITHUB_CLIENT_ID, VITE_BREVO_MA_KEY
npm run dev

The frontend runs on port 5173 by default and serves all four portals (Recruiter, Job Seeker, Developer, Admin) from a single Vite dev server.

API Reference

Resume Parsing

curl -X POST "https://api.vishleshan.ai/api/v1/parse" \
  -H "X-API-Key: vish_live_xxxxxxxxxxx" \
  -F "file=@resume.pdf"

{
  "success": true,
  "data": {
    "candidate_id": "cnd_9248239a",
    "name": "John Doe",
    "email": "johndoe@email.com",
    "skills": ["Distributed Systems", "Go", "Python"],
    "experience_years": 4.5
  }
}

Fraud Detection Scan

curl -X POST "https://api.vishleshan.ai/api/v1/protection/scan" \
  -H "X-API-Key: vish_live_xxxxxxxxxxx" \
  -d '{
    "scan_type": "job",
    "job_title": "Senior Frontend Engineer",
    "job_description": "We are looking for a React developer..."
  }'

{
  "success": true,
  "data": {
    "job_title": "Senior Frontend Engineer",
    "company_name": "Google",
    "originality_score": 94,
    "ai_probability": 6,
    "plagiarism_score": 5,
    "status": "Approved",
    "risk_level": "Low",
    "verified_company": "Yes",
    "flags": ["Source: LinkedIn"],
    "detailed_checks": {
      "official_website": { "status": "Yes", "details": "Official domain and secure certificates verified." },
      "recruiter_email": { "status": "Yes", "details": "Recruiter email domain matches company domain." },
      "salary_realistic": { "status": "Yes", "details": "Compensation aligns with market standards." },
      "linkedin_presence": { "status": "Yes", "details": "Found active company page on LinkedIn." },
      "description_copied": { "status": "No", "details": "Job requirements are custom-tailored." },
      "repeated_posts": { "status": "No", "details": "No duplicate posting signatures found." }
    },
    "summary": "Verification complete. Job listing appears safe and authentic."
  }
}

Rate Limits by Plan

Tier	Parses/mo	Match Ops/mo	Chat Queries/mo	Safety Scans/mo
Free	100	50	20	0
Starter	1,000	500	200	100
Business	10,000	Unlimited	Unlimited	1,000
Enterprise	Unlimited	Unlimited	Unlimited	Unlimited

Tech Stack

Layer	Technology
Frontend	React 18, Vite 8, Tailwind CSS 3, Zustand, Recharts, Framer Motion, GSAP, Radix UI
Backend	Django 5, Celery, PostgreSQL (Neon), Redis, ChromaDB
AI/LLM	Google Gemini (multi-key rotation), Custom 12+ agent pipeline
Payments	Razorpay
SMS & Email	2Factor (OTP), Brevo/Sendinblue (Transactional Email)
Auth	JWT + API Key + Google OAuth + GitHub OAuth
Deployment	Render (Backend), Vercel (Frontend), Neon (Database)

Project Structure

.
├── backend/
│   ├── agents/               # Multi-agent LLM system (12+ agents)
│   │   ├── parsing_agent.py
│   │   ├── normalization_agent.py
│   │   ├── matching_agent.py
│   │   ├── fraud_agent.py
│   │   ├── interview_agent.py
│   │   ├── resume_enhancer_agent.py
│   │   ├── salary_prediction_agent.py
│   │   └── llm.py            # LLM Router with multi-key rotation
│   ├── api/
│   │   ├── models.py          # Django ORM models
│   │   ├── views/
│   │   │   ├── admin_views.py     # Admin dashboard & moderation
│   │   │   ├── recruiter_auth.py  # Recruiter auth & profile
│   │   │   ├── seeker_auth.py     # Job seeker auth & profile
│   │   │   ├── seeker_jobs.py     # Job search & applications
│   │   │   ├── seeker_resume.py   # Resume management
│   │   │   ├── seeker_resume_builder.py  # AI resume builder
│   │   │   ├── protection.py     # Fraud detection endpoints
│   │   │   ├── round_views.py    # MCQ/Coding/Interview rounds
│   │   │   ├── sessions.py       # Recruitment session management
│   │   │   ├── developer/        # Developer portal APIs
│   │   │   ├── verification.py   # OTP & email verification
│   │   │   └── ...
│   │   ├── services/
│   │   │   ├── email_service.py       # Brevo email templates
│   │   │   ├── twofactor_service.py   # SMS OTP via 2Factor
│   │   │   ├── brevo_service.py       # Brevo CRM & tracking
│   │   │   └── notification_service.py
│   │   ├── decorators.py      # Auth, rate limiting, ban checks
│   │   ├── middleware.py      # CORS, usage logging
│   │   └── urls.py            # URL routing (25+ route groups)
│   ├── workers/               # Celery async task workers
│   ├── tests/                 # Unit & integration tests
│   ├── models/                # ML models & schemas
│   └── requirements.txt
├── frontend/
│   └── src/
│       ├── components/        # 47+ shared UI components
│       ├── pages/
│       │   ├── admin/         # Admin dashboard & login
│       │   ├── developer/     # Developer portal (12 pages)
│       │   ├── seeker/        # Seeker-specific pages
│       │   ├── user/          # Job seeker user pages (14 pages)
│       │   ├── test/          # Assessment round pages
│       │   ├── public/        # About, Terms, Contact, Refund
│       │   └── ...            # Landing, Auth, Dashboard, etc.
│       ├── stores/            # Zustand state management
│       └── lib/               # API clients & utilities
├── assets/                    # Screenshots & design assets
├── SECURITY.md                # Security policy & disclosure
├── CONTRIBUTING.md            # Contribution guidelines
├── LICENSE                    # MIT License
└── README.md

Security

Vishleshan includes comprehensive security controls:

IDOR Protection — Enforces company/seeker ownership checks on all candidate management views
JWT Blacklisting — Revokes tokens instantly upon logout using Redis-based blacklist with auto-expiration
Redis Ban Status Caching — 300s TTL cache with graceful DB fallback when Redis is unavailable
Atomic Cache Invalidation — Ban/unban operations clear Redis cache via transaction.on_commit() to prevent stale windows
Combined Rate Limiting — IP + Email rate limiting on admin login and support ticket endpoints with masked identifiers in 429 responses
Admin Self-Ban Prevention — Case-insensitive email matching against ADMIN_EMAIL prevents accidental admin lockout
Short Admin Token Expiry — Admin JWT tokens expire in 20 minutes
Audit Logging — AdminBanLog records every moderation action with admin email, target, and timestamp
XSS Protection — HTML template variables escaped at render-time (not storage) using django.utils.html.escape
Content Injection Protection — ReportLab PDF text inputs recursively HTML-escaped
Error Sanitization — Production responses mask internal details, returning generic correlation IDs
LLM Key Rotation — RotateLLMClient distributes across multiple Gemini API keys to prevent single-key exposure
CORS Configuration — Restricted allowed origins in production

Warning

Credential Rotation: All API keys, passwords, and connection URIs must be configured through .env files and are excluded from version control. If deploying to production, immediately rotate all previously used API keys that may persist in Git commit history.

For the full security policy, vulnerability reporting instructions, and severity classifications, see SECURITY.md.

License & Attributions

This project is licensed under the MIT License — see the LICENSE file for details.

Academic Context

Built as a Sem-IV Project at DAIICT | Multi-Agent Recruitment Intelligence Platform

Engineered for optimal performance, efficiency, zero-downtime operation, and seamless enterprise integration.

Name		Name	Last commit message	Last commit date
.agents		.agents
.github		.github
assets		assets
backend		backend
frontend		frontend
inspiration_ui		inspiration_ui
testing		testing
.gitignore		.gitignore
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
README.md		README.md
SECURITY.md		SECURITY.md
dummy.pdf		dummy.pdf
project-viva-prep.html		project-viva-prep.html
project_documentation.html		project_documentation.html
run.bat		run.bat
stillwater_hero.html		stillwater_hero.html

Latest commit

History

Folders and files

Repository files navigation

Vishleshan — Multi-Agent Recruitment Intelligence Platform

Overview

Multi-Agent System

Features

Recruiter Dashboard (ATS)

Fraud Detection & Protection

Job Seeker Portal

Admin Dashboard

Developer Portal (SaaS API)

Candidate Assessment Rounds

Architecture

backend/ — The AI Engine

frontend/ — The React SPA

Platform Highlights

1. Interactive Recruiter ATS (Applicant Tracking System)

2. Recruiter Recruitment Sessions

3. Interactive AI Resume Builder & Editor

4. Job Seeker Dashboard & ATS Scoring

5. Developer API Portal (SaaS Dashboard)

6. Job Seeker Landing & Smart Job Discovery

7. Market Trends & Salary Insights

Quick Start

One-Click Development Start (Windows)

Manual Setup & Requirements

1. Requirements

2. Backend Setup

3. Frontend Setup

API Reference

Resume Parsing

Fraud Detection Scan

Rate Limits by Plan

Tech Stack

Project Structure

Security

License & Attributions

Academic Context

About

Topics

Resources

Contributing

Security policy

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages

`backend/` — The AI Engine

`frontend/` — The React SPA