BharatMinds

Title: AI-Powered Intelligent Government Job Portal – SarkariSmart Team: BharatMinds | Track: AI/ML Problem: The government job ecosystem in India is highly fragmented. Crores of candidates visit multiple platforms daily - SSC, UPSC, IBPS, and state portals - searching for sarkari naukri, but there is no single intelligent system that aggregates all opportunities. Job notifications are released as lengthy PDFs that take hours to understand. Existing platforms fail to analyze a candidate’s skills, education, or behavior to suggest relevant roles. Meanwhile, government HR teams spend significant time manually entering job details from PDFs into portals. The National Career Service (NCS) also lacks AI-based recommendations, strong verification, fraud detection, and personalization. Proposed Solution – SarkariSmart - India's AI-powered Government Job Portal : A mobile-first AI-powered government job portal that connects verified organizations with qualified candidates through five intelligent layers: 1. Double Verification System: Only verified government organizations can post jobs. Registration requires a PAN number and an official government email (@gov.in/@nic.in), followed by manual admin approval. This removes fake job postings at the source. 2. AI PDF Parser: Organizations upload official job notification PDFs. Tesseract OCR extracts text, and the Sumy NLP library summarizes it. Job forms are auto-filled, reducing posting time from hours to minutes. 3. Three-Layer AI Recommendation Engine: Layer 1 – Weighted Scoring: Matches candidates to jobs using education (30%), stream (25%), skills (35%), and location (10%), generating a match percentage. Layer 2 – SVD Collaborative Filtering: Suggests jobs based on behavior—applications, likes, bookmarks, and dislikes. Layer 3 – Semantic Skill Matching: Uses TF-IDF and cosine similarity to understand skill relationships (e.g., “Python” matches “Django/Flask” roles). 4. AI Fraud Detection: AI analyzes job posts for suspicious keywords, unrealistic salaries, and missing details. Each job is assigned a trust badge: GREEN (Safe), YELLOW (Warning), RED (Danger). 5. Self-Learning Feedback Loop: User interactions train the system continuously. The SVD model retrains nightly, improving recommendations over time. 3-Role System: Employee: Personalized job recommendations, search/filter options, bookmarking, applications with cover letters, status tracking, profile management, courses, and news. Employer: PDF-based job posting, applicant management, status updates, and analytics dashboard. Admin: Organization verification, platform control, analytics, and fraud flag overrides. Tech Stack: Mobile: Expo React Native (Android + iOS) Backend: Django + Django REST Framework + JWT Auth + Rate Limiting Database: SQLite (dev), PostgreSQL (prod) ML Service: Python Flask (Scikit-learn, Surprise SVD, Tesseract OCR, Sumy NLP) APIs: ImgBB, NewsAPI, SMTP.js (all free) Total API Cost: ₹0 Key Improvements Over Existing Portals: 1. No AI in NCS → SarkariSmart uses a 3-layer AI recommendation system 2. Weak verification → Double verification with PAN, govt email, and admin approval 3. No fraud detection → AI-based trust scoring 4. No mobile-first experience → Dedicated React Native app 5. No feedback learning → Self-improving SVD model 6. Limited analytics → Role-based dashboards Impact: Candidates save hours daily in job search. Government HR teams reduce posting effort by up to 90%. A verified ecosystem minimizes fake or misleading listings, while AI continuously improves user experience. Future Scope: DigiLocker integration for document verification, AI resume analyzer, interview scheduling, regional language support, blockchain-based certificate validation, and integration with SSC/UPSC/IBPS results.