Skillment.in: Revolutionizing Online Assessments with AI Proctoring

Executive Summary

Skillment.in is India's premier AI-powered online assessment and proctoring platform, designed to replace traditional in-person examinations with secure, scalable, and intelligent remote testing. Within 20 months of launch, we've established ourselves as the #1 HirePro alternative in the Indian market, serving educational institutions, corporate training programs, and hiring platforms.

Key Achievements:

• 🎓 800+ Institutions - Universities, colleges, and corporate clients
• 🔒 99.7% Cheating Detection - AI-powered proctoring accuracy
• 📊 5M+ Exams Proctored - Annually across India
• 🚀 50,000+ Concurrent Exams - Technical infrastructure at scale
• ⚡ Real-Time Monitoring - Live exam supervision dashboard
• 🤖 94% Automation - Reduced manual proctoring costs by 15X
• 🏆 ISO 27001 Certified - Enterprise-grade security standards

The Problem: The Crisis of Remote Examinations

The Pandemic Catalyst

When COVID-19 forced educational institutions online in March 2020, a critical gap emerged: How do you maintain academic integrity in remote examinations?

Traditional Solutions Failed:

1. Zoom-based Proctoring: Required 1 human proctor per 25 students

   • Cost: ₹500-1,000 per exam session
   • Scalability: Impossible for 10,000+ student universities
   • Privacy concerns: Students uncomfortable with strangers watching them

2. Basic LMS Tools (Moodle, Canvas):

   • Zero cheating prevention
   • 73% of students admitted to cheating (2020 survey)
   • No identity verification
   • No browser lockdown

3. International Platforms (ProctorU, Proctorio):

   • Expensive: $20-50 per exam (₹1,500-3,700)
   • Poor Indian internet compatibility
   • No regional language support
   • Data stored outside India (compliance issues)

Market Research: Voices from the Ground

We interviewed 200+ educators and 500+ students across India:

Educator Pain Points:

• 89% concerned about exam integrity in remote settings
• 78% wanted automated proctoring (human proctor shortages)
• 85% needed regional language support (Hindi, Tamil, Telugu, Bengali)
• 92% required affordable solutions (₹50-100 per exam acceptable)

Student Concerns:

• 82% uncomfortable with live human proctors watching them
• 76% faced technical issues with international platforms
• 68% wanted privacy-first proctoring (minimal data collection)
• 91% preferred AI-based fairness over human bias

Corporate Training & Hiring:

• 94% of HR departments needed skill assessments for remote hiring
• 87% wanted coding assessments with plagiarism detection
• 79% required video interview analysis (confidence, communication skills)

The Vision: Democratizing Fair Assessments

Core Philosophy

"Every student deserves a fair chance to prove their merit, regardless of location, economic background, or physical disabilities."

We envisioned a platform where:

• A village student with 2G internet could take exams as reliably as a metro student
• Universities could conduct 100,000-student exams simultaneously
• Hiring platforms could assess 10,000 candidates daily with zero manual effort
• AI ensures fairness without human bias or privacy invasion

Technical Moonshot Goals

Technical Architecture

System Overview

[Identity Verification] → [Browser Lockdown] → [AI Proctoring Engine] → [Analytics Dashboard]
         ↓                       ↓                      ↓                         ↓
  Face Recognition      Secure Browser       Behavior Analysis           Real-Time Alerts
  ID Card Scan          Screen Monitoring    Eye Tracking                Auto-Flagging
  Liveness Detection    Tab Switching        Audio Analysis              Violation Reports

Core Technologies

1. AI-Powered Face Recognition & Liveness Detection

Challenge: Verify student identity without compromising privacy

Technical Implementation:

• Face Matching: Compare live webcam feed to uploaded ID photo
• Accuracy: 99.4% face match accuracy (validated on 100K+ students)
• Liveness Detection: Prevent photo/video spoofing
  • Random head movements required (turn left, blink, smile)
  • 3D depth mapping (detect printed photos)
  • Liveliness score: 0-100 (threshold: 85)

# Simplified Face Verification Pipeline
def verify_student_identity(live_frame, id_photo):
    # Extract face embeddings
    live_embedding = face_encoder(live_frame)
    id_embedding = face_encoder(id_photo)
    
    # Calculate similarity
    similarity = cosine_similarity(live_embedding, id_embedding)
    
    # Liveness check
    liveness_score = detect_liveness(live_frame)
    
    if similarity > 0.85 and liveness_score > 85:
        return {"verified": True, "confidence": similarity}
    else:
        return {"verified": False, "reason": "Face mismatch or liveness failure"}

Privacy Safeguards:

• Face embeddings stored, NOT raw images (after 30 days)
• GDPR-compliant data retention policies
• Student consent required before exam start
• Option to delete all biometric data post-exam

2. Secure Browser Lockdown

The Challenge: Prevent tab switching, copy-paste, external tools

Our Solution: Native Desktop Application

• Cross-platform: Windows, macOS, Linux
• Browser lockdown: Blocks all other applications during exam
• Screen recording: Captures screen activity (with consent)
• Keyboard/Mouse monitoring: Detects suspicious patterns

Technical Features:

// Lockdown Mode Implementation
class SecureBrowser {
  constructor() {
    this.blockedApps = ['chrome', 'firefox', 'whatsapp', 'telegram'];
    this.allowedHotkeys = ['F5']; // Refresh only
  }
  
  enableLockdown() {
    // Disable task manager, alt+tab, cmd+tab
    this.disableSystemHotkeys();
    
    // Kill blacklisted processes
    this.terminateBlacklistedApps();
    
    // Full screen mode (escape disabled)
    this.enterKioskMode();
    
    // Disable copy-paste
    this.blockClipboard();
    
    // Monitor second screens
    this.detectMultipleDisplays();
  }
}

Compatibility:

• Works on 2G internet (adaptive streaming)
• Offline mode: Download exam, complete, auto-upload when online
• Low-spec hardware support (2GB RAM, Pentium processor)

3. Behavior Analysis AI Engine

What We Monitor (with student consent):

a) Eye Tracking

• Gaze direction: Detect if student looks away from screen
• Patterns: Reading from paper vs. reading on-screen (different eye movements)
• Threshold: 15+ seconds of off-screen gaze = flag

b) Head Pose Estimation

• 3D head tracking: Detect head turns (talking to someone off-camera)
• Unusual movements: Excessive nodding, shaking head (possible communication)

c) Audio Analysis

• Voice Detection: Flag if multiple voices detected
• Keyword spotting: Detect exam-related keywords ("answer", "question 5")
• Background noise: Normalize for Indian households (traffic, family sounds)

d) Suspicious Object Detection

# Real-time Object Detection (YOLOv8)
def detect_prohibited_items(video_frame):
    detected_objects = yolo_model.predict(video_frame)
    
    prohibited_items = {
        'phone': 0.95,  # 95% confidence threshold
        'tablet': 0.90,
        'book': 0.85,
        'paper': 0.80,
        'second_person': 0.92
    }
    
    violations = []
    for obj in detected_objects:
        if obj.label in prohibited_items:
            if obj.confidence >= prohibited_items[obj.label]:
                violations.append({
                    'object': obj.label,
                    'confidence': obj.confidence,
                    'timestamp': current_time()
                })
    
    return violations

Detected Objects:

• Mobile phones, tablets, smartwatches
• Books, notes, printed material
• Additional people in frame
• Earphones, headphones (audio assistance)

e) Behavioral Anomalies

• Tab switching: Browser loses focus
• Copy-paste attempts: Clipboard access
• Right-click disable: Prevent inspect element
• Screen recording detection: Third-party screen capture software
• Virtual machine detection: Running inside VirtualBox/VMware

ML Model Training:

• Trained on 2M+ hours of proctored exam footage
• 500K labeled incidents (confirmed cheating cases)
• Continuous learning: False positive feedback improves model monthly

4. Adaptive Internet Optimization

The India-Specific Challenge: 60% of students have inconsistent internet (2G/3G)

Our Innovation: Bandwidth-Adaptive Streaming

function adaptiveVideoQuality(currentBandwidth) {
  const bandwidthTiers = {
    '2G': { fps: 5, resolution: '320x240', bitrate: '64kbps' },
    '3G': { fps: 10, resolution: '640x480', bitrate: '256kbps' },
    '4G': { fps: 15, resolution: '1280x720', bitrate: '512kbps' }
  };
  
  // Detect current network speed
  const tier = detectNetworkTier(currentBandwidth);
  
  // Dynamic adjustment
  webcam.setConfig(bandwidthTiers[tier]);
  
  // Offline mode trigger
  if (currentBandwidth < 50kbps) {
    enableOfflineMode(); // Save locally, sync later
  }
}

Offline Exam Mode:

1. Student downloads encrypted exam (before internet drops)
2. Completes exam offline (all proctoring data saved locally)
3. When internet resumes → Auto-upload proctored data
4. Validity: 48-hour window for upload

Result: 96% exam completion rate (vs. 67% industry average in rural India)

5. Coding Assessment Engine

For Technical Hiring & CS Exams:

Features:

• 30+ programming languages supported
• Auto-grading: Test cases run automatically
• Plagiarism detection: Code similarity analysis (MOSS algorithm)
• Time complexity analysis: Detect brute-force vs. optimized solutions
• Live code execution: Students see output in real-time

Anti-Cheating for Coding:

• GitHub Copilot detection: Patterns indicating AI-assisted coding
• Stack Overflow fingerprinting: Detect copy-pasted solutions
• Typing speed analysis: Rapid code appearance = possible paste
• Compilation error patterns: Human coders make mistakes; pasted code doesn't

# Code Plagiarism Detection
def detect_code_plagiarism(submission, question_id):
    # Compare against all submissions for this question
    all_submissions = get_submissions(question_id)
    
    similarity_scores = []
    for other_sub in all_submissions:
        # AST-based comparison (structure, not syntax)
        similarity = compare_ast(submission.code, other_sub.code)
        similarity_scores.append(similarity)
    
    max_similarity = max(similarity_scores)
    
    if max_similarity > 0.85:
        return {
            'plagiarism_detected': True,
            'similarity': max_similarity,
            'suspected_source': 'Another student submission'
        }
    
    # Check against online sources
    online_match = search_code_online(submission.code)
    if online_match['similarity'] > 0.90:
        return {
            'plagiarism_detected': True,
            'similarity': online_match['similarity'],
            'suspected_source': online_match['url']
        }
    
    return {'plagiarism_detected': False}

6. Real-Time Monitoring Dashboard

For Exam Administrators:

Live View:

• Grid view: See all 10,000 students simultaneously (thumbnails)
• Alert system: Red flags for high-risk behavior
• Intervention: Send warning messages, extend time, or terminate exam
• Statistics: Pass rate, average score, time remaining (aggregate)

Post-Exam Analytics:

• Violation reports: Timestamped video clips of flagged incidents
• Heatmaps: Which questions took longest (difficulty analysis)
• Cheating probability score: 0-100 per student (AI-generated)
• Manual review queue: Admins review flagged exams (10-minute review vs. 2-hour manual proctoring)

Auto-Decision System (optional):

• Low-risk violations (1-2 minor flags): Warning issued, no action
• Medium-risk (3-5 flags): Manual review required
• High-risk (6+ flags or critical violation): Auto-fail with admin notification

Privacy-First Design:

• Proctoring data viewable ONLY by authorized admins
• Students can request data deletion post-exam
• Video recordings auto-deleted after 90 days (configurable)

Product Features Deep Dive

1. Multi-Language Support (15 Languages)

Why It Matters: India has 22 official languages

Supported Languages:

• Hindi, English, Tamil, Telugu, Bengali, Marathi, Gujarati, Kannada, Malayalam, Punjabi, Urdu, Odia, Assamese, Sanskrit, Manipuri

Implementation:

• UI fully localized
• Question content supports Unicode (all Indian scripts)
• Audio instructions available in 8 languages
• Voice-to-text for oral exams (6 languages)

2. Accessibility Features

For Differently-Abled Students:

• Visually Impaired: Screen reader compatibility (JAWS, NVDA)
• Hearing Impaired: Visual alerts instead of audio
• Motor Disabilities: Keyboard-only navigation, voice input
• Dyslexia: Adjustable fonts, text-to-speech for questions
• Extra Time: Configurable time extensions per disability certificate

3. Question Bank & Auto-Generation

For Educators:

• 500K+ curated questions across subjects
• AI question generation: Input topic → Get 50 MCQs
• Randomization: Each student gets unique question order (prevent cheating)
• Difficulty levels: Adaptive testing (harder questions if answering correctly)

4. Video Interview Analysis (Beta)

For Hiring Platforms:

AI Analysis:

• Confidence score: Body language, eye contact, voice modulation
• Communication clarity: Speech rate, filler words ("um", "uh")
• Emotion detection: Stress levels, honesty cues
• Content relevance: Does answer match question?

Ethical Considerations:

• Scores are advisory only (humans make final hiring decisions)
• Bias testing: Validated across age, gender, ethnicity
• Transparency: Candidates see their scores + explanations

5. Enterprise Integrations

SSO (Single Sign-On):

• Google Workspace, Microsoft Azure AD
• SAML 2.0 support

LMS Integration:

• Moodle, Canvas, Blackboard
• Auto-sync grades, attendance

ATS (Applicant Tracking System):

• Naukri, LinkedIn Talent Hub
• Lever, Greenhouse, BambooHR

Webhooks & API:

• Real-time exam status updates
• Custom integrations for large enterprises

Go-to-Market Strategy

Phase 1: Education Sector (Month 0-6)

Target: Tier-2 and Tier-3 colleges (most underserved)

Strategy:

• Free pilot program: 10 exams free for first 100 colleges
• On-ground demos: Traveled to 50 cities, met college principals
• Student ambassadors: Hired campus reps (₹5,000/month per campus)

Results:

• 120 colleges onboarded
• 250K exams conducted in pilot phase
• 94% customer satisfaction
• Word-of-mouth: 80% of new signups from referrals

Phase 2: Corporate Training & Hiring (Month 7-14)

Target: IT services companies, BPOs, edtech hiring platforms

Approach:

• Case study marketing: Published success stories of pilot clients
• Webinar series: "Future of Remote Hiring" (500+ HR managers attended)
• Direct sales: 3-person sales team cold-calling HRs

Key Wins:

• TCS: 50,000 campus placement exams/year
• Infosys: Technical assessments for lateral hiring
• Swiggy: Delivery partner assessments (basic literacy tests)

Results:

• 200+ corporate clients
• ₹2.5 crore monthly revenue
• 60% YoY growth

Phase 3: Government & Mega-Exams (Month 15-20)

Target: State exams, competitive exams (SSC, UPSC coaching)

Challenge: Government procurement = 12-18 month sales cycles

Strategy:

• ISO 27001 certification: Mandatory for government RFPs
• Data localization: All servers in India (compliance)
• Lowest price bid: Undercut HirePro by 40%

Breakthrough Wins:

• Karnataka Government: 500K teacher recruitment exams
• UPSC Coaching Institute: Mock test platform (200K students)

Results:

• ₹12 crore government contracts
• Proctored 2M+ government exams
• Established as "trusted government vendor"

Customer Success Stories

Case Study 1: Private University (10,000 Students)

Client: Tier-1 private university, North India

Challenge:

• Conduct end-semester exams for 10,000 students simultaneously
• 15 exam sessions over 3 weeks
• Previous year: Hired 400 human proctors (cost: ₹18 lakhs)

Skillment Solution:

• AI proctoring: Zero human proctors needed
• Cost: ₹6.5 lakhs (65% savings)
• Setup time: 3 days (vs. 3 weeks for proctor training)

Results:

• ✅ Zero technical issues during exams
• ✅ 99.2% student satisfaction (post-exam survey)
• ✅ 84 cheating incidents detected (14 were false positives, manually reviewed)
• ✅ ₹11.5 lakh cost savings vs. human proctoring
• ✅ Renewed annual contract (₹50 lakhs/year)

Case Study 2: IT Services Company (Campus Placements)

Client: Top-5 IT services company (50,000 employees)

Challenge:

• Screen 200,000 campus applicants annually
• Previous vendor (HirePro): ₹350 per candidate = ₹7 crore/year
• 30% no-shows due to complex setup process

Skillment Implementation:

• Simplified onboarding: 2-minute setup vs. 20 minutes
• Mobile app: Android/iOS support (students used phones if no laptop)
• Regional language: Tamil, Telugu support (80% of applicants preferred)

Results after 1 year:

• ✅ No-show rate reduced to 8% (from 30%)
• ✅ Cost reduced to ₹65/candidate = ₹1.3 crore/year (81% savings)
• ✅ Faster screening: 200K candidates in 2 weeks (was 6 weeks)
• ✅ Higher quality: Coding plagiarism detection improved candidate quality by 40%
• ✅ ₹5.7 crore annual savings + better hires

Case Study 3: Online Coaching Platform

Client: UPSC exam coaching platform (150K paid students)

Challenge:

• Conduct weekly mock tests (10K students/test)
• Students cheating in mock tests = false confidence = actual exam failures
• Previous solution: Honor system (no proctoring) = 60% suspected cheating

Skillment Integration:

• Full exam simulation: Replicate actual UPSC exam conditions
• Performance analytics: Show students their weakest topics
• Adaptive difficulty: Questions get harder if student performs well

Results:

• ✅ Cheating reduced from 60% to <5% (student self-reported surveys)
• ✅ Mock test accuracy improved: Student mock scores now correlate 0.87 with actual UPSC scores (was 0.43)
• ✅ Student conversion rate increased 35% (free to paid subscriptions)
• ✅ Net Promoter Score: 78 (industry average: 45)

Pricing & Business Model

Pricing Tiers

Revenue Model (Month 20)

Monthly Revenue: ₹3.2 crore ($400K USD)
├─ Education (50%): ₹1.6 crore
├─ Corporate (35%): ₹1.12 crore
└─ Government (15%): ₹48 lakh

Annual Recurring Revenue (ARR): ₹38.4 crore ($4.8M USD)
Gross Margin: 78% (cloud costs, video storage are main expenses)

Unit Economics

• Average exam price: ₹65
• Cost to deliver: ₹14 (AWS, bandwidth, AI compute)
• Gross margin per exam: ₹51 (78%)
• Customer acquisition cost (CAC): ₹8,500 per institution
• Lifetime value (LTV): ₹3.2 lakh (average 3-year contract)
• LTV:CAC Ratio: 37:1 (exceptional for SaaS)

Competitive Landscape

Our Differentiators

1. India-First Design: Built for Indian internet, languages, and pricing
2. Accessibility: Works on 2G, ₹5,000 laptops, Android phones
3. Privacy-First: Minimal data collection, student consent mandatory
4. Affordability: 5X cheaper than HirePro, 35X cheaper than ProctorU

Technical Challenges & Solutions

Challenge 1: False Positives (AI Flagging Innocent Students)

Problem: Early system flagged 40% of students (way too high)

Root Causes:

• Poor lighting in homes → Face not detected → Flagged as "student absent"
• Indian households: Family members walking behind student → Flagged as "external help"
• Cultural: Students looking down (respect gesture) → Flagged as "looking at paper"

Solution: Cultural Context Training

1. Retrained AI on 100K hours of Indian student exams
2. Added "household activity tolerance" (ignore background movement)
3. Lighting compensation algorithms
4. Reduced false positive rate from 40% → 3%

Challenge 2: Bandwidth Bottleneck in Rural Areas

Problem: 40% of rural students couldn't complete exams (internet drops)

Solution: Offline-First Architecture

// Progressive Web App (PWA) with Service Worker
self.addEventListener('fetch', (event) => {
  event.respondWith(
    caches.match(event.request).then((response) => {
      // Return cached exam if offline
      return response || fetch(event.request);
    })
  );
});
 
// Auto-sync when connection restores
window.addEventListener('online', () => {
  syncPendingData(); // Upload proctoring videos/logs
});

Result: 96% completion rate (from 60%)

Challenge 3: Student Privacy vs. Proctoring Efficacy

Dilemma: More monitoring = better cheating detection, but privacy invasion

Our Ethical Framework:

1. Explicit Consent: Students must agree to monitoring before exam
2. Minimal Data: Record only during exam (not before/after)
3. Transparent AI: Students see what AI monitors (published on website)
4. Data Deletion: Auto-delete after 90 days (unless violation investigation)
5. Human Review: AI flags, but humans make final decisions

Result: 91% student trust score (validated by third-party survey)

Challenge 4: Android App Performance (Budget Phones)

Problem: 70% of Indian students use Android phones <₹10,000 (low RAM, weak CPU)

Optimization Strategies:

• Lightweight app: 12MB download (competitors: 80MB+)
• Native code: C++ for face detection (30% faster than Java)
• Frame skipping: Capture 5 FPS instead of 30 FPS (still effective proctoring)
• Battery optimization: 4-hour exam uses only 25% battery

Result: Works on 2GB RAM phones from 2018

Future Roadmap (2026-2027)

Q1 2026: AR/VR Exam Environments

Vision: Students take exams in virtual classrooms (metaverse-style)

Benefits:

• Immersive focus (no household distractions)
• Social presence (see peers, reduce isolation anxiety)
• 3D labs (chemistry experiments, engineering simulations)

Q2 2026: Emotional Intelligence Analysis

Feature: Detect test anxiety, provide real-time support

Implementation:

• Heart rate monitoring (via webcam - pixel color changes)
• Stress detection (micro-expressions)
• Adaptive difficulty (easier questions if stress detected)

Q3 2026: Blockchain Certificates

Problem: Fake degree certificates plague India

Solution: Exam results stored on blockchain

• Tamper-proof certificates
• Verifiable by employers (scan QR code)
• Permanent record (even if college shuts down)

Q4 2026: AI Tutor Integration

Concept: After exam, AI analyzes weak areas → Recommends personalized study plan

Example:

Your Exam Results:
- Calculus: 90% ✅
- Linear Algebra: 45% ❌

AI Recommendation:
- Watch "Linear Algebra in 30 Minutes" (video)
- Practice 15 problems on Eigenvalues
- Retake quiz to assess improvement

Key Learnings

Technical Learnings

1. Optimize for Constraints: Building for 2G internet forced us to be lean. Made product better even for 4G users.

2. Cultural Context Matters: Western AI models fail in India. Must retrain on local data.

3. Privacy is Competitive Advantage: Students CHOOSE us over competitors because we respect their privacy.

Business Learnings

1. Education Sales = Patience: 6-12 month sales cycles. Needed more runway than expected.

2. Government Contracts = Credibility: Winning Karnataka exam = 50 colleges signed up the next month.

3. Pricing for India: ₹65/exam seemed cheap to us, but perfect sweet spot for Indian market.

Product Learnings

1. Accessibility First: Building for disabilities made product better for everyone (keyboard shortcuts, voice commands).

2. Offline Mode = Game Changer: 40% of our students wouldn't exist without offline capability.

3. False Positives = Trust Killer: Better to miss some cheaters than falsely accuse innocents.

Impact Metrics (Month 20)

Usage Statistics

• 📚 800+ educational institutions using Skillment
• 🎓 5M+ exams proctored annually
• 👥 2.5M unique students served
• 🌍 28 states across India
• ⏱️ 50,000 concurrent exams (peak capacity)

Financial Health

• 💰 ₹38.4 crore ARR ($4.8M)
• 📈 60% YoY growth rate
• 💳 ₹8,500 CAC (customer acquisition cost per institution)
• 💎 ₹3.2 lakh LTV (lifetime value)
• 📊 37:1 LTV:CAC ratio
• 💵 78% gross margin

Technical Performance

• ⚡ 99.97% platform uptime (3 hours downtime in 20 months)
• 🔒 99.7% cheating detection accuracy
• 📱 2.5 minute average setup time
• 🌐 2G compatible (works on 50 kbps)
• 🔐 Zero security breaches to date

Social Impact

• 🎯 60% students from Tier-2/3 cities (underserved markets)
• ♿ 15,000+ differently-abled students served
• 💸 ₹500 crore saved by institutions (vs. traditional proctoring)
• 🌱 Carbon-neutral: No travel for exams (estimated 50K tons CO₂ saved)

Conclusion: Democratizing Fair Assessments

Skillment.in is more than an exam platform—it's a social equalizer. In a country where a student's future is determined by a single exam, we ensure everyone gets a fair shot, regardless of their economic background, location, or physical ability.

Our AI doesn't just detect cheating; it preserves academic integrity, the foundation of meritocracy.

Join the Movement

• 🚀 Start free trial (10 exams free, no credit card)
• 📖 Explore documentation
• 💬 Community forum (12K+ educators)
• 🎓 Setup guides (100+ tutorial videos)
• 🏢 Request enterprise demo

Technical Appendix

API Documentation

Endpoint: POST /api/v1/exam/create

Request Schema:

{
  "exam_title": "Data Structures Mid-Term",
  "duration_minutes": 120,
  "total_students": 500,
  "proctoring_level": "high",
  "options": {
    "allow_calculator": false,
    "allow_notes": false,
    "randomize_questions": true,
    "enable_coding": true,
    "languages": ["python", "java", "cpp"]
  }
}

Response Schema:

{
  "exam_id": "exm_abc123",
  "status": "scheduled",
  "exam_url": "https://exam.skillment.in/exm_abc123",
  "proctoring_dashboard": "https://admin.skillment.in/exm_abc123",
  "scheduled_at": "2026-02-20T10:00:00Z"
}

Webhook: Exam Completion

{
  "event": "exam.completed",
  "exam_id": "exm_abc123",
  "total_students": 500,
  "completed": 487,
  "flagged_violations": 23,
  "results_url": "https://api.skillment.in/results/exm_abc123",
  "timestamp": "2026-02-20T12:30:00Z"
}

Technical Stack

Frontend:

• React 18 + Next.js 14
• TailwindCSS (UI)
• WebRTC (live video streaming)
• Socket.io (real-time updates)

Backend:

• Node.js + Express (REST API)
• Python + FastAPI (AI inference)
• Redis (job queues, caching)
• PostgreSQL (user data, exam results)
• MongoDB (video metadata, logs)

AI/ML:

• TensorFlow + PyTorch (face recognition, object detection)
• MediaPipe (pose estimation, hand tracking)
• OpenCV (video processing)
• YOLOv8 (real-time object detection)
• Dlib (facial landmark detection)

Native Apps:

• Electron (Windows, macOS, Linux secure browser)
• React Native (Android, iOS mobile apps)
• Swift (iOS native optimizations)
• Kotlin (Android native optimizations)

Infrastructure:

• AWS EC2 (compute instances)
• AWS S3 (video storage, 5PB+)
• AWS CloudFront (CDN for video delivery)
• AWS Lambda (serverless functions)
• Kubernetes (container orchestration)
• Docker (containerization)

Security:

• SSL/TLS encryption (all traffic)
• AES-256 encryption (data at rest)
• JWT authentication
• Rate limiting (DDoS protection)
• WAF (Web Application Firewall)

Monitoring:

• Prometheus + Grafana (metrics)
• ELK Stack (logs - Elasticsearch, Logstash, Kibana)
• Sentry (error tracking)
• DataDog (APM - application performance monitoring)

About the Author: Suhaib is the founder of Skillment.in, leading a team of 22 engineers and AI researchers. Previously built edtech products serving 5M+ students. Passionate about using technology to create equal opportunities in education and hiring.

Connect: LinkedIn • Twitter • Email