Pratinidhi: AI-Powered Educational Communication Platform Based on Intelligent Message Processing and Eligibility Filtering
K Bhanu Srinija¹, Macha Praveen², Sesha Amrutha Madhavarapu³, Sivalanka Pranay⁴
¹²³⁴ Department of Computer Science and Engineering (AI & ML)
Aditya College of Engineering and Technology (Autonomous), Surampalem, AP - 533437
Guide: Mrs. Vijaya Nirmala, M.Tech., (Ph.D.), Associate Professor
Abstract
Educational institutions face persistent challenges in disseminating placement announcements and academic notifications through unstructured channels such as WhatsApp groups, causing students to miss critical opportunities due to information overload. This paper presents Pratinidhi, an AI-powered educational communication platform that addresses this challenge through a novel two-stage message processing architecture using Google Gemini 2.0 Flash AI. The V0 stage extracts concise summaries, message type classifications, structured eligibility criteria (CGPA, backlogs, gender, batch year, branch), deadlines, and key action points. The V1 stage captures comprehensive details including company information, roles, compensation packages, application procedures, and contact details. A rule-based eligibility checking algorithm compares student profiles against V0-extracted criteria across six dimensions, ensuring students receive V1 details only when they meet all requirements. The platform implements four-tier role-based access control (Super Admin, Institution Admin, Teacher, Student) with hierarchical approval workflows and real-time messaging via Supabase Realtime WebSocket. Built with Next.js 16, React 19, Supabase PostgreSQL, and TypeScript, the system achieves 98% V0 accuracy (2.3s) and 95% V1 accuracy (3.1s), reducing information noise by 60%. Validated through 125 Vitest unit tests and 57 integration tests, the platform is deployed at prosody-v3.vercel.app.
Keywords: AI-Powered Communication, Eligibility Filtering, Google Gemini AI, Large Language Model, Message Processing, Educational Technology
