The rapid growth of remote collaboration has made online meeting platforms an essential component of modern academic and enterprise workflows. Existing systems such as Google Meet, Zoom, Microsoft Teams, and Calendly focus primarily on connectivity, scheduling, and screen sharing, but offer no intelligent mechanism to help users rehearse, evaluate, or improve their communication before real meetings. This paper presents an AI-Powered Meeting Scheduler with a Smart Meeting Twin, a unified platform that combines classical scheduling and video conferencing with a persona-based conversational AI trainer. The Smart Meeting Twin engages users in voice-first interviews across HR, Technical, Professor, CEO, Client, and Team Lead personas using large language models, real-time speech-to-text, and neural text-tospeech pipelines. After each session, the platform analyzes the candidate's transcript to produce quantitative scores for confidence, communication, fluency, grammar, and pronunciation, together with narrative strengths, weaknesses, and personalized suggestions. The underlying architecture uses WebRTC for peer-to-peer media, WebSocket signaling, a REST API gateway, an authentication layer backed by JSON Web Tokens, and a managed PostgreSQL database with row-level security. This paper reviews recent literature on video conferencing, AI-assisted scheduling, natural language processing, and virtual meeting assistants published between 2021 and 2026. A comparative analysis with existing tools identifies significant research gaps, particularly the absence of integrated interview simulation, confidence estimation, and personalized coaching within scheduling platforms. We propose a modular design framework covering user management, meeting orchestration, the Smart Meeting Twin engine, and analytics reporting. Preliminary evaluation on a prototype shows that the system produces stable low-latency conversations and consistent, explainable performance reports. The paper concludes with a discussion of practical advantages, deployment considerations, and future directions including emotion detection, AI avatars, multilingual support, and automatic minutes of meeting generation. The proposed system contributes an original, self-contained architecture that unifies scheduling, live collaboration, and AIdriven communication coaching within a single privacy-aware platform suitable for academic and professional adoption.

Dr.A.Somasundaram, Dharineesh.M, "AI-Powered Meeting Scheduler with Smart Meeting Twin" International Journal of Computer Science and Engineering Development, V1(3): Page(13-21) July-August 2026. ISSN: 3139-0862. www.ijcsed.com. Published by Scientific and Academic Research Publishing.